H1@V-^m;4Wg<&0T*E43hX&L&p$$qDprKhvt+--jT7}7np#A3
zem<@ulZcFPQ@L2!n>{z** Rsngu)}*?GTJhe4@nkM;9p
zo0~XgelYr}k! MD;#w>^Lk=pU~xA+05pL5C+nuprD;sTQ156_CxpGyRv^KE+gXj?#it8lG9c!
z&L**=B<#YM;R6n8YLNNoOvTyS^xFCkEIpTMYgR%igz|=mxblhF$
- 1.2. [Framework of Precision Medicine Process](#12)
+
+1.1. [Definition](#11)
+
+1.2. [Framework of Precision Medicine Process](#12)
+
2. [Deep Phenotyping](#2)
- 2.1. [Importance of Phenotype](#21)
- 2.2. [Why "Deep" Phenotyping?](#22)
- 2.3. [Processing Deep Phenotyping Data](#23)
+
+2.1. [Importance of Phenotype](#21)
+
+2.2. [Why "Deep" Phenotyping?](#22)
+
+2.3. [Processing Deep Phenotyping Data](#23)
+
3. [Data Analysis](#3)
- 3.1. [Preprocessing and Data Mining](#31)
- 3.2. [Diagnostic and Prognostic Models](#32)
- 3.3. [Predicting Treatment Response](#33)
+
+3.1. [Preprocessing and Data Mining](#31)
+
+3.2. [Diagnostic and Prognostic Models](#32)
+
+3.3. [Predicting Treatment Response](#33)
+
4. [Evolving Precision Medicine](#4)
+
5. [Conclusion](#5)
+
6. [References](#6)
+
+
+
## 1. What is Precision Medicine?
+
+
### 1) Definition
+
+
Precision medicine is personalized treatment strategies on the basis of genetic, biomarker, phenotypic or psychosocial characteristics to stratify patients into novel subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. It is also commonly referred to as 'stratified medicine', 'targeted therapy' and 'deep phenotyping.'
+
+
The focus of applied definitions is moving away from classical 'signs-and-symptoms’ approach or 'population’ approach into 'N of 1’ approach in which each patient is an entire trial as a single case study from N number of individuals within a population. Figure 1 shows the difference between two approaches with same group of patients. Comprehensive study of such subclasses ultimately depends on computational resources to capture, store and exchange phenotypic data, and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.
+
+

-##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and give the same treatment, because they were treated as one group from the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group are divided into disease subgroups with more precise and validated phenotypic recognition or better understanding of the causal pathways.
+
+
+##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and give the same treatment, because they were treated as one group from the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group are divided into disease subgroups with more precise and validated phenotypic recognition or better understanding of the causal pathways.
+
+
### 2) Framework of Precision Medicine Process
-The framework of precision medicine is a process made up of number of feedback loops, with no steady-end point. Each cycle is an attempt to make the process more precise with more stratification of patients. The data assessed in the patients are used to try to develop clinically relevant models, and results of these analyses then inform the further assessment of patients as an evolving result.
+
+
+The framework of precision medicine is a process made up of number of feedback loops, with no steady-end point. Each cycle is an attempt to make the process more precise with more stratification of patients. The data assessed in the patients are used to try to develop clinically relevant models, and results of these analyses then inform the further assessment of patients as an evolving result.
+
+

-###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from track 1-3 are fed back to deep phenotyping stage for subsequent assessments.
+
+
+###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from track 1-3 are fed back to deep phenotyping stage for subsequent assessments.
+
+
## 2. Deep Phenotyping
+
### 1) Importance of Phenotype
+
+
The word phenotype can be used differently in biology versus medicine. In biology, ‘phenotype’ is collection of observable physical properties of an organism, including the organism's appearance, development, behavior, and even characteristics such as gene expression profiles in response to environmental cues. In medical context, however, ‘phenotype’ is deviation from normal morphology, physiology, or behavior. When physician makes note of patient’s phenotype, the physician is doing so by taking medical history, performing physical examination, diagnostic imaging, blood tests, and other psychological tests.
-This is where precise phenotype information comes into play. As physician makes note of patient’s phenotypes, the physician is making a diagnosis for the patient’s disease – making a hypothesis of what it may or may not be and provides treatment which may or may not work. Making a diagnosis, therefore, is perhaps the most important task in treating a disease, but it is the most challenging task especially for the cases of rare diseases of which 8000 diseases are yet to be classified. Today, major clinical problems arise from delayed or inaccurate diagnosis, treatment and unnecessary procedures that result in patients’ psychological burdens and unnecessary expenses. To prevent complications and set forth effective treatments from making correct prognosis, full spectrum of phenotypic abnormalities is absolutely critical for improving care quality while reducing the need for unnecessary diagnostic testing and therapies. Figure 3 shows importance of phenotype and its molecular network underpinning. Even with the same endophenotype, risk varies between individuals, because their phenotype ultimately varies in the molecular level.
+
+
+This is where precise phenotype information comes into play. As physician makes note of patient’s phenotypes, the physician is making a diagnosis for the patient’s disease – making a hypothesis of what it may or may not be and provides treatment which may or may not work. Making a diagnosis, therefore, is perhaps the most important task in treating a disease, but it is the most challenging task especially for the cases of rare diseases of which 8000 diseases are yet to be classified. Today, major clinical problems arise from delayed or inaccurate diagnosis, treatment and unnecessary procedures that result in patients’ psychological burdens and unnecessary expenses. To prevent complications and set forth effective treatments from making correct prognosis, full spectrum of phenotypic abnormalities is absolutely critical for improving care quality while reducing the need for unnecessary diagnostic testing and therapies. Figure 3 shows importance of phenotype and its molecular network underpinning. Even with the same endophenotype, risk varies between individuals, because their phenotype ultimately varies in the molecular level.
+
+

+
+
##### Figure 3: Individuals with the same endophenotype, such as hypertension, can have different molecular phenotypes – resulting in different risk levels. Molecular phenotyping determines patients have different treatment responses as well.
+
+
### 2) Why "Deep" Phenotyping?
-The major problem with current phenotyping regime is sloppy or imprecise descriptions of phenotypic descriptions in medical publications. For instance, “still walking after 25 years of onset” is definitely a phenotypic description, yet it is unclear what the physician observed in the patient as such descriptions are likely to evoke different ideas for different readers depending on personal experience and imagination. Furthermore, the current gene mutation database has very little to no phenotypic information. Other than the fact that the patient has that mutagen. While this information is indeed important for determining pathogenicity, it is devoid of natural history of the disease – which is very important clinically in understanding the spectrum of complications with the disease, or genotype-phenotype correlations. Deep phenotyping offers solutions for current challenges, including semantics and technical standards for phenotype and disease data. Deep phenotyping is defined as precise and comprehensive analysis of phenotypic abnormalities in which individual components of individual components of the phenotype are observed and described. Figure 4 shows precision medicine approach to phenotyping; even the individuals with same endophenotype may are biologically distinct and encompass different disease profiles.
+
+
+The major problem with current phenotyping regime is sloppy or imprecise descriptions of phenotypic descriptions in medical publications. For instance, “still walking after 25 years of onset” is definitely a phenotypic description, yet it is unclear what the physician observed in the patient as such descriptions are likely to evoke different ideas for different readers depending on personal experience and imagination. Furthermore, the current gene mutation database has very little to no phenotypic information. Other than the fact that the patient has that mutagen. While this information is indeed important for determining pathogenicity, it is devoid of natural history of the disease – which is very important clinically in understanding the spectrum of complications with the disease, or genotype-phenotype correlations. Deep phenotyping offers solutions for current challenges, including semantics and technical standards for phenotype and disease data. Deep phenotyping is defined as precise and comprehensive analysis of phenotypic abnormalities in which individual components of individual components of the phenotype are observed and described. Figure 4 shows precision medicine approach to phenotyping; even the individuals with same endophenotype may are biologically distinct and encompass different disease profiles.
+
+

+
+
##### Figure 4: Individuals undergo deep phenotyping process with data analysis performed using network analysis. This is deep phenotyping, because individuals are classified beyond their endophenotypes. Such methodology optimizes identification of biomarkers, clinical trials, and prognostication of disease.
+
+
### 3) Processing Deep Phenotyping Data
-Conversion of deep phenotyping data into tangible therapeutic utility poses number of challenges yet to be solved. First, computational environments for analysis of high dimensional data is required. Second, there must be data from large populations of patients as a knowledge network. The most important, however, is the integration of the two. For example, next-Generation sequencing enables genome-wide investigation of rare genetic variants, but associating these with diseases requires tailored statical tools that summarize the information of neighbored variants. Indeed, the integration of two types of different information are critical in generating the data in clinically relevant way. To clarify this, the structure of the precision medicine process offers tracks 1-3, which are handling of the data for better prediction of drug sensitivity.
+
+
+Conversion of deep phenotyping data into tangible therapeutic utility poses number of challenges yet to be solved. First, computational environments for analysis of high dimensional data is required. Second, there must be data from large populations of patients as a knowledge network. The most important, however, is the integration of the two. For example, next-Generation sequencing enables genome-wide investigation of rare genetic variants, but associating these with diseases requires tailored statical tools that summarize the information of neighbored variants. Indeed, the integration of two types of different information are critical in generating the data in clinically relevant way. To clarify this, the structure of the precision medicine process offers tracks 1-3, which are handling of the data for better prediction of drug sensitivity.
+
+
+
## 3. Data Analysis
-### 1)
-### 2)
+
+Large-scale of data is expected to be integrated and converted into more precise therapeutic interventions. The analysis of deep phenotyping is distinguished into three sequential tracks: in track 1, the data are handled without knowledge of a clinical end-point; in track 2, data are used to build models for a more precise diagnosis or prognosis of disease or disease outcome; and track 3 leads to models that predict more precisely how well specific patients respond to treatment. [Afzal]
+
+
+
+##### Figure 5: An integrated precision medicine framework for heterogeneous data collection, data analysis, knowledge management, and implementation of knowledge and data services.
+
+### 1) Preprocessing and Data Mining
+
+The first step of data analysis is always data preprocessing. Studies on fractional exhaled nitric oxide as a marker of eosinophilic inflammation in patients with asthma might be influenced by variable techniques of measurement, sampling procedures, breathing manoeuvres and different types of devices. Therefore, data is expected to be filtered and adjusted to be comparable in analysis. **Quality Control** and **Preprocessing** of the data shall be preformed in this step. Since quality control is always dependent on the specific data type, more details can be found in paper. [Ferté ]
+
+Some of the common approaches for data preprocessing includes:
+
+1. **Denoising** and Baseline Correction
+
+2. Missing Value Filling
+
+3. Data Out of Range Filtration
+
+4. Data Below the Limits of Detection Filtration
+
+5. **Variable Selection**
+
+
+**Case Example: Denoising + Variable selection**
+
+*The collection of deep phenotyping data usually includes great amount of unselected data, which will inevitably include variables that are irrelevant for later modelling. The irrelevant variables that contain random “noise” can mask important underlying relationships and structures, which can shown in below.*
+
+45 individuals are clustered into three homogeneous subgroups if two variables are considered. A hierarchical cluster analysis using only these two variables identifies three clusters clearly (figure 6b). However, if three random noise variables are added to the dataset, the cluster algorithm fails to find the three groups as indicated in figure 6c. We can see that about half of the individuals are classified incorrectly.
+
+
+
+##### Figure 6: Impact of variable selection based on illustration
+
+Variable selection is important for filtering out noise data, yet how to define noise or whether a variable is revevant is always a unanswerable question. One important factor that can guide variable selection is scientifically based *a priori* knowledge about the possibly **biologically relevant variables**.
+
+### 2) Diagnostic and Prognostic Models
+
+After data is ready from track 1 for further training, models can be developed to estimate the risk with absolute probability of the presence or absence of an outcome or disease in individuals based on their clinical and non-clinical information. Prediction model can be diagnostic (outcome or disease present at this moment) or prognostic (outcome occurs within a specified time frame). [Hendriksen]
+
+
+
+##### Figure 7: Representation of diagnostic and prognostic prediction modeling studies
+
+- Model Development
+Two main strategies are usually used to develop the models:
+
+ 1) **Full Model**
+ No predictor selection is applied . Pros: avoid improper predictor selection due to predictor selection bias. Cons: requires much prior knowledge to adequately preselect the biologically relevant predictors for modelling.
+
+ 2) **Variable selection**
+ **Backward Elimination** of ‘redundant’ predictors or **Forward Selection** of ‘promising’ ones. The backward procedure is initialized with the full multivariable and then subsequently removes predictors based on a predefined criterion. Forward selection is when predictors are added to the multivariable model one by one. There is no agreement on the optimal method among the two methods, The choice of methods is highly *Content-Specific*.
+
+- Outcome
+The outcome of a prediction aims to reflects a clinically significant and patient relevant health state, for example, death yes or no, or absence or presence of Leukemia. In case of prognostic prediction model, a follow-up period is needed to be clearly defined for outcome development. [Hendriksen]
+
+- Clinics-friendly Accommodation
+Regression model can be too complicated to use in daily clinical uses, the model is usually simplify by rounding coefficient toward integer numbers for easier scoring. However, such accommodation might negatively effected the accuracy of the model and thus needs to be applied carefully.
+
+
### 3) Predicting Treatment Response
+
+
After developing diagnostic and prognostic models from the previous step, there is a need for assessing variables that define a novel taxonomy with their relevance in predicting treatment response. There are two strategies to develop models that predict treatment response.
+
+
- Prediction models can be built on the diagnostic and prognostic models from the previous stage as prognostic factors may act as the natural variables to consider when developing prediction models. For instance, epidermal growth factor receptor tyrosine kinase status acts as a prognostic factor for survival in patients with non-small cell lung cancer and a predictive factor for response to the tyrosine kinase inhibitor gefitinib as first-line treatment at the same time [Riley RD(will change later after sorting&numbering sources].
+
- The data can be directly utilized to extract significant information relevant to the prediction of the disease. For instance, in the first clinical trials that used the monoclonal anti-IL-5 antibody mepolizumab for asthma treatment, the use of mepolizumab was associated with a significant reduction blood and sputum eosinophils but did not have significant clinical benefit in asthma patients [Flood-Page]. Consequently, in the following trials, patients with refractory eosinophilic asthma were selected and in this subgroup, mepolizumab therapy added significant clinical benefit in patients by reducing exacerbations and improving asthma quality of life scores [Haldar P].
+
+

+
##### Figure 5: The figure above is an illustration of applying different treatments, also known as "tailored medicine", to categorized groups to increase the effectiveness of treatment based on a prediction model.
-This process of building prediction models involves generating further knowledge about the disease and treatment from the diagnotics and prognostics models or the data itself. The findings from this process, in the form of data, would be fed back to the phenotyping of patients to further adjust the clinical trial that could more precisely show the effectiveness of the treatment.
+
+
+This process of building prediction models involves generating further knowledge about the disease and treatment from the diagnotics and prognostics models or the data itself. The findings from this process, in the form of data, would be fed back to the phenotyping of patients to further adjust the clinical trial that could more precisely show the effectiveness of the treatment.
+
+
In addition to the feedback, dissemination and communication of the taxonomy and models with the clinical and scientific communities, for instance, to provide utilizable algorithms for clinical practices.
+
+
## 4. Evolving Precision Medicine
+
+
##### Figure 6: The figure below shows repeated cycles of precision medicine and its subsequent outcome.
+

-As seen in the figure, the cycle of patients assessment(deep phenotyping), data processing(preprocessing and data mining), and model building(diagnotic, prognostic, and prediction model building) is repeated at least several times to further increase the preciseness of the medicine by categorizing patient groups at a higher resolution.
+
+
+As seen in the figure, the cycle of patients assessment(deep phenotyping), data processing(preprocessing and data mining), and model building(diagnotic, prognostic, and prediction model building) is repeated at least several times to further increase the preciseness of the medicine by categorizing patient groups at a higher resolution.
+
+
The detailed steps are as follows:
+
+
1. In the first cycles, patients are categorized into diagnostic/prognostic groups based on obvious characteristics.
-2. In later cycles, more specific groups of patients are defined using more in-depth data that was obtained from the previous cycle as the data from each cycle is fed back to the next cycle of patient assessment.
+
+2. In later cycles, more specific groups of patients are defined using more in-depth data that was obtained from the previous cycle as the data from each cycle is fed back to the next cycle of patient assessment.
+
3. Eventually, these feedback loops may allow the final cycles to target individual patients with specific data profiles.
+
+
## 5. Conclusion
-In this paper, the process of precision medicine, starting from the deep phenotyping stage to data analysis part, which comprised preprocessing, data mining, developing diagnostic, prognostic, and prediction models, were discussed against the background of airway diseases. As seen in the previous section on the continual evolution of precision medicine with repeated cycle of the above steps, precision medicine should be viewed as the continuous process of feedback loops rather than a steady-state with an end-point or a specific output from the research. In this context, tailored or stratified medicine resulting from new stratifications can be considered as a makeshift product of the process that will generate new data for the next cycle, thus further increasing the precision.
+
+
+In this paper, the process of precision medicine, starting from the deep phenotyping stage to data analysis part, which comprised preprocessing, data mining, developing diagnostic, prognostic, and prediction models, were discussed against the background of airway diseases. As seen in the previous section on the continual evolution of precision medicine with repeated cycle of the above steps, precision medicine should be viewed as the continuous process of feedback loops rather than a steady-state with an end-point or a specific output from the research. In this context, tailored or stratified medicine resulting from new stratifications can be considered as a makeshift product of the process that will generate new data for the next cycle, thus further increasing the precision.
+
+
Once again, the ultimate goal of precision medicine is the most effective treatment for the individual patient. To step closer to this goal, ongoing endeavors to be even more precise and individualistic, newly-gained scientific and clinical knowledge, and new data sources would be necessary.
+
+
## 6. References
+
Inke R. König, Oliver Fuchs, Gesine Hansen, Erika von Mutius, Matthias V. Kopp
-European Respiratory Journal Oct 2017, 50 (4) 1700391; DOI: 10.1183/13993003.00391-2017
+
+European Respiratory Journal Oct 2017, 50 (4) 1700391; DOI: 10.1183/13993003.00391-2017
+
+
Ginsburg GS, Phillips KA. Precision Medicine: From Science To Value. Health Aff (Millwood). 2018 May;37(5):694-701. doi: 10.1377/hlthaff.2017.1624. PMID: 29733705; PMCID: PMC5989714.
+
+
Robinson PN. Deep phenotyping for precision medicine. Hum Mutat. 2012 May;33(5):777-80. doi: 10.1002/humu.22080. PMID: 22504886.
+
+
National Research Council, Committee on A Framework for Developing a New Taxonomy of Disease. Toward Precision Medicine: Building a Knowledge Network for Biomedical Research and a New Taxonomy of Disease. National Academies Press; Washington DC: 2011.
+
+
Leopold JA, Loscalzo J. Emerging Role of Precision Medicine in Cardiovascular Disease. Circ Res. 2018;122(9):1302-1315. doi:10.1161/CIRCRESAHA.117.310782
+
+
Riley RD, Hayden JA, Steyerberg EW, et al. Prognosis Research Strategy (PROGRESS) 2: prognostic factor research. PLoS Med 2013; 10: e1001380.
+
+
Flood-Page P, Swenson C, Faiferman I, et al. A study to evaluate safety and efficacy of mepolizumab in patients with moderate persistent asthma. Am J Respir Crit Care Med 2007; 176: 1062–1071.
+
+
Haldar P, Brightling CE, Hargadon B, et al. Mepolizumab and exacerbations of refractory eosinophilic asthma. N Engl J Med 2009; 360: 973–984.
-the image- might not use-- “What Is Precision Medicine and How Can It Help Fix Healthcare.” ReferralMD, ReferralMD, 11 Dec. 2018, getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/.
+
+
+the image- might not use-- “What Is Precision Medicine and How Can It Help Fix Healthcare.” ReferralMD, ReferralMD, 11 Dec. 2018, getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/.
+
https://getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/
+
+ Ferté C, Trister AD, Huang E, et al. Impact of bioinformatic procedures in the development and translation of high-throughput molecular classifiers in oncology. Clin Cancer Res 2013; **19**: 4315–4325.
+
+Hendriksen JMT, Geersing GJ, Moons KGM, de Groot JAH. Diagnostic and prognostic prediction models. J Thromb Hae- most 2013; 11 (Suppl. 1): 129–41.
+
+M. Afzal, S. M. Riazul Islam, M. Hussain and S. Lee, "Precision Medicine Informatics: Principles, Prospects, and Challenges," in IEEE Access, vol. 8, pp. 13593-13612, 2020, doi: 10.1109/ACCESS.2020.2965955.
+
From 2d28b7472cfe6b62d559184a76221c5c9499ad7f Mon Sep 17 00:00:00 2001
From: meixwu
@@ -53,19 +50,14 @@
Precision medicine is personalized treatment strategies on the basis of genetic, biomarker, phenotypic or psychosocial characteristics to stratify patients into novel subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. It is also commonly referred to as 'stratified medicine', 'targeted therapy' and 'deep phenotyping.'
-<<<<<<< HEAD
The focus of applied definitions is moving away from classical 'signs-and-symptoms’ approach or 'population’ approach into 'N of 1’ approach in which each patient is an entire trial as a single case study from N number of individuals within a population. Figure 1 shows the difference between two approaches with same group of patients. Comprehensive study of such subclasses ultimately depends on computational resources to capture, store and exchange phenotypic data, and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.
-=======
-The focus is to move away from classical 'signs-and-symptoms’ approach or 'population’ approach into 'N of 1’ approach in which each patient is an entire trial or a single case study from N number of individuals within a population. Figure 1 shows the difference between classical approach and precision medicine approach. Ultimately, comprehensive study of such subclasses depends on computational resources to capture, store and exchange phenotypic data and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.
->>>>>>> d94479fce00c00580bbf6152cab074d85af22e2f

-<<<<<<< HEAD
##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and give the same treatment, because they were treated as one group from the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group are divided into disease subgroups with more precise and validated phenotypic recognition or better understanding of the causal pathways.
@@ -87,17 +79,6 @@ The framework of precision medicine is a process made up of number of feedback l
###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from track 1-3 are fed back to deep phenotyping stage for subsequent assessments.
-=======
-##### Figure 1: The left diagram shows treatment based on 'signs-and-symptoms' research. Patients with colon cancer are clustered into one group and given the same therapy, which thereby varies in treatment responses. Treatment has worked for some individuals while it had no or adverse effect for other. This is because this approach does not take patients' phenoytpic information into account. On the right, the same group of patients have been divided into subgroups based on individual phenotypic recognition, resulting in better understanding of the causal pathways and better treatment responses.
-
-### 2) Framework of Precision Medicine Process
-
-At the highest level, precision medicine process is a framework made up of feedback loops, with no steady-end point. Each cycle is an attempt to sort out and validate the process and stratify the subgroups further. The data assessed in the patients are used to try to develop clinically relevant models, and analyses inform need for further assessment of patients as an evolving result.
-
-
-
-###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from track 1-3 are fed back to deep phenotyping stage for subsequent assessments.
->>>>>>> d94479fce00c00580bbf6152cab074d85af22e2f
## 2. Deep Phenotyping
From 9a918b5cb2c03803ca6871ca3b711bcfd170a5bc Mon Sep 17 00:00:00 2001
From: meixwu
-1.1. [Definition](#11)
+ 1.1. [Definition](#11)
-1.2. [Framework of Precision Medicine Process](#12)
+ 1.2. [Framework of Precision Medicine Process](#12)
2. [Deep Phenotyping](#2)
-2.1. [Importance of Phenotype](#21)
+ 2.1. [Importance of Phenotype](#21)
-2.2. [Why "Deep" Phenotyping?](#22)
+ 2.2. [Why "Deep" Phenotyping?](#22)
-2.3. [Processing Deep Phenotyping Data](#23)
+ 2.3. [Processing Deep Phenotyping Data](#23)
3. [Data Analysis](#3)
-3.1. [Preprocessing and Data Mining](#31)
+ 3.1. [Preprocessing and Data Mining](#31)
-3.2. [Diagnostic and Prognostic Models](#32)
+ 3.2. [Diagnostic and Prognostic Models](#32)
-3.3. [Predicting Treatment Response](#33)
+ 3.3. [Predicting Treatment Response](#33)
4. [Evolving Precision Medicine](#4)
@@ -135,7 +131,7 @@ Large-scale of data is expected to be integrated and converted into more precise
##### Figure 5: An integrated precision medicine framework for heterogeneous data collection, data analysis, knowledge management, and implementation of knowledge and data services.
-### 1) Preprocessing and Data Mining
+### 1) Preprocessing and Data Mining
The first step of data analysis is always data preprocessing. Studies on fractional exhaled nitric oxide as a marker of eosinophilic inflammation in patients with asthma might be influenced by variable techniques of measurement, sampling procedures, breathing manoeuvres and different types of devices. Therefore, data is expected to be filtered and adjusted to be comparable in analysis. **Quality Control** and **Preprocessing** of the data shall be preformed in this step. Since quality control is always dependent on the specific data type, more details can be found in paper. [Ferté ]
@@ -167,7 +163,7 @@ Some of the common approaches for data preprocessing includes:
Variable selection is important for filtering out noise data, yet how to define noise or whether a variable is revevant is always a unanswerable question. One important factor that can guide variable selection is scientifically based *a priori* knowledge about the possibly **biologically relevant variables**.
-### 2) Diagnostic and Prognostic Models
+### 2) Diagnostic and Prognostic Models
After data is ready from track 1 for further training, models can be developed to estimate the risk with absolute probability of the presence or absence of an outcome or disease in individuals based on their clinical and non-clinical information. Prediction model can be diagnostic (outcome or disease present at this moment) or prognostic (outcome occurs within a specified time frame). [Hendriksen]
From 56b643d166d79b1f8b4ca1119f5753c33e82827c Mon Sep 17 00:00:00 2001
From: Soyeon Kim <37919264+ksoyeona@users.noreply.github.com>
Date: Wed, 16 Dec 2020 08:57:21 +0900
Subject: [PATCH 59/94] Update Precision_Medicine.md
---
.../Precision_Medicine.md | 117 +-----------------
1 file changed, 5 insertions(+), 112 deletions(-)
diff --git a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
index 1f91372..68046d4 100644
--- a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
+++ b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
@@ -1,29 +1,20 @@
# The Process of Precision Medicine
### by Soyeon Kim, Hwayeon Lee, Meixian Wu
-
-
-
+
1. [What is Precision Medicine?](#1)
-
1.1. [Definition](#11)
-
1.2. [Framework of Precision Medicine Process](#12)
-
+
2. [Deep Phenotyping](#2)
-
2.1. [Importance of Phenotype](#21)
-
2.2. [Why "Deep" Phenotyping?](#22)
-
2.3. [Processing Deep Phenotyping Data](#23)
3. [Data Analysis](#3)
3.1. [Preprocessing and Data Mining](#31)
-
3.2. [Diagnostic and Prognostic Models](#32)
-
3.3. [Predicting Treatment Response](#33)
4. [Evolving Precision Medicine](#4)
@@ -32,97 +23,50 @@
6. [References](#6)
-
-
-
-
## 1. What is Precision Medicine?
-
-
### 1) Definition
-
-
Precision medicine is personalized treatment strategies on the basis of genetic, biomarker, phenotypic or psychosocial characteristics to stratify patients into novel subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. It is also commonly referred to as 'stratified medicine', 'targeted therapy' and 'deep phenotyping.'
-
-
The focus of applied definitions is moving away from classical 'signs-and-symptoms’ approach or 'population’ approach into 'N of 1’ approach in which each patient is an entire trial as a single case study from N number of individuals within a population. Figure 1 shows the difference between two approaches with same group of patients. Comprehensive study of such subclasses ultimately depends on computational resources to capture, store and exchange phenotypic data, and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.
-
-

-
-
##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and give the same treatment, because they were treated as one group from the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group are divided into disease subgroups with more precise and validated phenotypic recognition or better understanding of the causal pathways.
-
-
### 2) Framework of Precision Medicine Process
-
-
The framework of precision medicine is a process made up of number of feedback loops, with no steady-end point. Each cycle is an attempt to make the process more precise with more stratification of patients. The data assessed in the patients are used to try to develop clinically relevant models, and results of these analyses then inform the further assessment of patients as an evolving result.
-
-

-
-
###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from track 1-3 are fed back to deep phenotyping stage for subsequent assessments.
-
-
## 2. Deep Phenotyping
### 1) Importance of Phenotype
-
-
The word phenotype can be used differently in biology versus medicine. In biology, ‘phenotype’ is collection of observable physical properties of an organism, including the organism's appearance, development, behavior, and even characteristics such as gene expression profiles in response to environmental cues. In medical context, however, ‘phenotype’ is deviation from normal morphology, physiology, or behavior. When physician makes note of patient’s phenotype, the physician is doing so by taking medical history, performing physical examination, diagnostic imaging, blood tests, and other psychological tests.
-
-
This is where precise phenotype information comes into play. As physician makes note of patient’s phenotypes, the physician is making a diagnosis for the patient’s disease – making a hypothesis of what it may or may not be and provides treatment which may or may not work. Making a diagnosis, therefore, is perhaps the most important task in treating a disease, but it is the most challenging task especially for the cases of rare diseases of which 8000 diseases are yet to be classified. Today, major clinical problems arise from delayed or inaccurate diagnosis, treatment and unnecessary procedures that result in patients’ psychological burdens and unnecessary expenses. To prevent complications and set forth effective treatments from making correct prognosis, full spectrum of phenotypic abnormalities is absolutely critical for improving care quality while reducing the need for unnecessary diagnostic testing and therapies. Figure 3 shows importance of phenotype and its molecular network underpinning. Even with the same endophenotype, risk varies between individuals, because their phenotype ultimately varies in the molecular level.
-
-

-
-
##### Figure 3: Individuals with the same endophenotype, such as hypertension, can have different molecular phenotypes – resulting in different risk levels. Molecular phenotyping determines patients have different treatment responses as well.
-
-
### 2) Why "Deep" Phenotyping?
-
-
The major problem with current phenotyping regime is sloppy or imprecise descriptions of phenotypic descriptions in medical publications. For instance, “still walking after 25 years of onset” is definitely a phenotypic description, yet it is unclear what the physician observed in the patient as such descriptions are likely to evoke different ideas for different readers depending on personal experience and imagination. Furthermore, the current gene mutation database has very little to no phenotypic information. Other than the fact that the patient has that mutagen. While this information is indeed important for determining pathogenicity, it is devoid of natural history of the disease – which is very important clinically in understanding the spectrum of complications with the disease, or genotype-phenotype correlations. Deep phenotyping offers solutions for current challenges, including semantics and technical standards for phenotype and disease data. Deep phenotyping is defined as precise and comprehensive analysis of phenotypic abnormalities in which individual components of individual components of the phenotype are observed and described. Figure 4 shows precision medicine approach to phenotyping; even the individuals with same endophenotype may are biologically distinct and encompass different disease profiles.
-
-

-
-
##### Figure 4: Individuals undergo deep phenotyping process with data analysis performed using network analysis. This is deep phenotyping, because individuals are classified beyond their endophenotypes. Such methodology optimizes identification of biomarkers, clinical trials, and prognostication of disease.
-
-
### 3) Processing Deep Phenotyping Data
-
-
Conversion of deep phenotyping data into tangible therapeutic utility poses number of challenges yet to be solved. First, computational environments for analysis of high dimensional data is required. Second, there must be data from large populations of patients as a knowledge network. The most important, however, is the integration of the two. For example, next-Generation sequencing enables genome-wide investigation of rare genetic variants, but associating these with diseases requires tailored statical tools that summarize the information of neighbored variants. Indeed, the integration of two types of different information are critical in generating the data in clinically relevant way. To clarify this, the structure of the precision medicine process offers tracks 1-3, which are handling of the data for better prediction of drug sensitivity.
-
-
-
## 3. Data Analysis
Large-scale of data is expected to be integrated and converted into more precise therapeutic interventions. The analysis of deep phenotyping is distinguished into three sequential tracks: in track 1, the data are handled without knowledge of a clinical end-point; in track 2, data are used to build models for a more precise diagnosis or prognosis of disease or disease outcome; and track 3 leads to models that predict more precisely how well specific patients respond to treatment. [Afzal]
@@ -147,7 +91,6 @@ Some of the common approaches for data preprocessing includes:
5. **Variable Selection**
-
@@ -185,115 +128,65 @@ The outcome of a prediction aims to reflects a clinically significant and patien
- Clinics-friendly Accommodation
Regression model can be too complicated to use in daily clinical uses, the model is usually simplify by rounding coefficient toward integer numbers for easier scoring. However, such accommodation might negatively effected the accuracy of the model and thus needs to be applied carefully.
-
### 3) Predicting Treatment Response
-
-
After developing diagnostic and prognostic models from the previous step, there is a need for assessing variables that define a novel taxonomy with their relevance in predicting treatment response. There are two strategies to develop models that predict treatment response.
-
-
- Prediction models can be built on the diagnostic and prognostic models from the previous stage as prognostic factors may act as the natural variables to consider when developing prediction models. For instance, epidermal growth factor receptor tyrosine kinase status acts as a prognostic factor for survival in patients with non-small cell lung cancer and a predictive factor for response to the tyrosine kinase inhibitor gefitinib as first-line treatment at the same time [Riley RD(will change later after sorting&numbering sources].
-
- The data can be directly utilized to extract significant information relevant to the prediction of the disease. For instance, in the first clinical trials that used the monoclonal anti-IL-5 antibody mepolizumab for asthma treatment, the use of mepolizumab was associated with a significant reduction blood and sputum eosinophils but did not have significant clinical benefit in asthma patients [Flood-Page]. Consequently, in the following trials, patients with refractory eosinophilic asthma were selected and in this subgroup, mepolizumab therapy added significant clinical benefit in patients by reducing exacerbations and improving asthma quality of life scores [Haldar P].
-
-

##### Figure 8: The figure above is an illustration of applying different treatments, also known as "tailored medicine", to categorized groups to increase the effectiveness of treatment based on a prediction model.
-
-
This process of building prediction models involves generating further knowledge about the disease and treatment from the diagnotics and prognostics models or the data itself. The findings from this process, in the form of data, would be fed back to the phenotyping of patients to further adjust the clinical trial that could more precisely show the effectiveness of the treatment.
-
-
In addition to the feedback, dissemination and communication of the taxonomy and models with the clinical and scientific communities, for instance, to provide utilizable algorithms for clinical practices.
-
-
## 4. Evolving Precision Medicine
-
-
##### Figure 9: The figure below shows repeated cycles of precision medicine and its subsequent outcome.

-
-
As seen in the figure, the cycle of patients assessment(deep phenotyping), data processing(preprocessing and data mining), and model building(diagnotic, prognostic, and prediction model building) is repeated at least several times to further increase the preciseness of the medicine by categorizing patient groups at a higher resolution.
-
-
The detailed steps are as follows:
-
-
1. In the first cycles, patients are categorized into diagnostic/prognostic groups based on obvious characteristics.
2. In later cycles, more specific groups of patients are defined using more in-depth data that was obtained from the previous cycle as the data from each cycle is fed back to the next cycle of patient assessment.
3. Eventually, these feedback loops may allow the final cycles to target individual patients with specific data profiles.
-
-
## 5. Conclusion
-
-
In this paper, the process of precision medicine, starting from the deep phenotyping stage to data analysis part, which comprised preprocessing, data mining, developing diagnostic, prognostic, and prediction models, were discussed against the background of airway diseases. As seen in the previous section on the continual evolution of precision medicine with repeated cycle of the above steps, precision medicine should be viewed as the continuous process of feedback loops rather than a steady-state with an end-point or a specific output from the research. In this context, tailored or stratified medicine resulting from new stratifications can be considered as a makeshift product of the process that will generate new data for the next cycle, thus further increasing the precision.
-
-
Once again, the ultimate goal of precision medicine is the most effective treatment for the individual patient. To step closer to this goal, ongoing endeavors to be even more precise and individualistic, newly-gained scientific and clinical knowledge, and new data sources would be necessary.
-
-
## 6. References
-Inke R. König, Oliver Fuchs, Gesine Hansen, Erika von Mutius, Matthias V. Kopp
-
-European Respiratory Journal Oct 2017, 50 (4) 1700391; DOI: 10.1183/13993003.00391-2017
-
-
+Inke R. König, Oliver Fuchs, Gesine Hansen, Erika von Mutius, Matthias V. Kopp European Respiratory Journal Oct 2017, 50 (4) 1700391; DOI: 10.1183/13993003.00391-2017
Ginsburg GS, Phillips KA. Precision Medicine: From Science To Value. Health Aff (Millwood). 2018 May;37(5):694-701. doi: 10.1377/hlthaff.2017.1624. PMID: 29733705; PMCID: PMC5989714.
-
-
Robinson PN. Deep phenotyping for precision medicine. Hum Mutat. 2012 May;33(5):777-80. doi: 10.1002/humu.22080. PMID: 22504886.
-
-
National Research Council, Committee on A Framework for Developing a New Taxonomy of Disease. Toward Precision Medicine: Building a Knowledge Network for Biomedical Research and a New Taxonomy of Disease. National Academies Press; Washington DC: 2011.
-
-
Leopold JA, Loscalzo J. Emerging Role of Precision Medicine in Cardiovascular Disease. Circ Res. 2018;122(9):1302-1315. doi:10.1161/CIRCRESAHA.117.310782
-
-
Riley RD, Hayden JA, Steyerberg EW, et al. Prognosis Research Strategy (PROGRESS) 2: prognostic factor research. PLoS Med 2013; 10: e1001380.
-
-
Flood-Page P, Swenson C, Faiferman I, et al. A study to evaluate safety and efficacy of mepolizumab in patients with moderate persistent asthma. Am J Respir Crit Care Med 2007; 176: 1062–1071.
-
-
Haldar P, Brightling CE, Hargadon B, et al. Mepolizumab and exacerbations of refractory eosinophilic asthma. N Engl J Med 2009; 360: 973–984.
-
-
-the image- might not use-- “What Is Precision Medicine and How Can It Help Fix Healthcare.” ReferralMD, ReferralMD, 11 Dec. 2018, getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/.
-
-https://getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/
+the image- might not use-- “What Is Precision Medicine and How Can It Help Fix Healthcare.” ReferralMD, ReferralMD, 11 Dec. 2018, getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/. https://getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/
- Ferté C, Trister AD, Huang E, et al. Impact of bioinformatic procedures in the development and translation of high-throughput molecular classifiers in oncology. Clin Cancer Res 2013; **19**: 4315–4325.
+Ferté C, Trister AD, Huang E, et al. Impact of bioinformatic procedures in the development and translation of high-throughput molecular classifiers in oncology. Clin Cancer Res 2013; **19**: 4315–4325.
Hendriksen JMT, Geersing GJ, Moons KGM, de Groot JAH. Diagnostic and prognostic prediction models. J Thromb Hae- most 2013; 11 (Suppl. 1): 129–41.
From 4ff38e2bd8e5720b72f1c4be84a9948c087294a4 Mon Sep 17 00:00:00 2001
From: Soyeon Kim <37919264+ksoyeona@users.noreply.github.com>
Date: Wed, 16 Dec 2020 08:58:12 +0900
Subject: [PATCH 60/94] Update Precision_Medicine.md
---
finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md | 6 ------
1 file changed, 6 deletions(-)
diff --git a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
index 68046d4..9adb76f 100644
--- a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
+++ b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
@@ -5,22 +5,16 @@
1. [What is Precision Medicine?](#1)
1.1. [Definition](#11)
1.2. [Framework of Precision Medicine Process](#12)
-
2. [Deep Phenotyping](#2)
2.1. [Importance of Phenotype](#21)
2.2. [Why "Deep" Phenotyping?](#22)
2.3. [Processing Deep Phenotyping Data](#23)
-
3. [Data Analysis](#3)
-
3.1. [Preprocessing and Data Mining](#31)
3.2. [Diagnostic and Prognostic Models](#32)
3.3. [Predicting Treatment Response](#33)
-
4. [Evolving Precision Medicine](#4)
-
5. [Conclusion](#5)
-
6. [References](#6)
## 1. What is Precision Medicine?
From 9e67da332ed287bd01d5b26e9ca9b17fedf734e6 Mon Sep 17 00:00:00 2001
From: Soyeon Kim <37919264+ksoyeona@users.noreply.github.com>
Date: Wed, 16 Dec 2020 09:00:15 +0900
Subject: [PATCH 61/94] Update Precision_Medicine.md
---
finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md | 3 ---
1 file changed, 3 deletions(-)
diff --git a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
index 9adb76f..3e6d6d7 100644
--- a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
+++ b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
@@ -84,10 +84,7 @@ Some of the common approaches for data preprocessing includes:
4. Data Below the Limits of Detection Filtration
5. **Variable Selection**
-
-
-
**Case Example:** *Denoising + Variable selection*
*The collection of deep phenotyping data usually includes great amount of unselected data, which will inevitably include variables that are irrelevant for later modelling. The irrelevant variables that contain random “noise” can mask important underlying relationships and structures, which can shown in below.*
From 85d461323de106c694c55c1a72fdd0304af5d507 Mon Sep 17 00:00:00 2001
From: Soyeon Kim <37919264+ksoyeona@users.noreply.github.com>
Date: Wed, 16 Dec 2020 09:00:55 +0900
Subject: [PATCH 62/94] Update Precision_Medicine.md
---
finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md | 2 ++
1 file changed, 2 insertions(+)
diff --git a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
index 3e6d6d7..c0469ce 100644
--- a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
+++ b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
@@ -84,7 +84,9 @@ Some of the common approaches for data preprocessing includes:
4. Data Below the Limits of Detection Filtration
5. **Variable Selection**
+
+
**Case Example:** *Denoising + Variable selection*
*The collection of deep phenotyping data usually includes great amount of unselected data, which will inevitably include variables that are irrelevant for later modelling. The irrelevant variables that contain random “noise” can mask important underlying relationships and structures, which can shown in below.*
From 285653b3673d983603e6451151789eb066b36c5f Mon Sep 17 00:00:00 2001
From: Soyeon Kim <37919264+ksoyeona@users.noreply.github.com>
Date: Wed, 16 Dec 2020 09:26:35 +0900
Subject: [PATCH 63/94] Update Precision_Medicine.md
---
.../Precision_Medicine.md | 46 +++++++++----------
1 file changed, 23 insertions(+), 23 deletions(-)
diff --git a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
index c0469ce..1aae017 100644
--- a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
+++ b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
@@ -21,29 +21,29 @@
### 1) Definition
-Precision medicine is personalized treatment strategies on the basis of genetic, biomarker, phenotypic or psychosocial characteristics to stratify patients into novel subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. It is also commonly referred to as 'stratified medicine', 'targeted therapy' and 'deep phenotyping.'
+Precision medicine is personalized treatment strategies on the basis of genetic, biomarker, phenotypic or psychosocial characteristics to stratify patients into novel subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. It is also commonly referred to as 'stratified medicine', 'targeted therapy', and 'deep phenotyping.'
-The focus of applied definitions is moving away from classical 'signs-and-symptoms’ approach or 'population’ approach into 'N of 1’ approach in which each patient is an entire trial as a single case study from N number of individuals within a population. Figure 1 shows the difference between two approaches with same group of patients. Comprehensive study of such subclasses ultimately depends on computational resources to capture, store and exchange phenotypic data, and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.
+The focus of applied definitions is moving away from the classical 'signs-and-symptoms’ approach or 'population’ approach into the 'N of 1’ approach in which each patient is an entire trial as a single case study from N number of individuals within a population. Figure 1 shows the difference between two approaches with the same group of patients. A comprehensive study of such subclasses ultimately depends on computational resources to capture, store and exchange phenotypic data, and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.

-##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and give the same treatment, because they were treated as one group from the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group are divided into disease subgroups with more precise and validated phenotypic recognition or better understanding of the causal pathways.
+##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and give the same treatment, because they were treated as one group for the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group is divided into disease subgroups with more precise and validated phenotypic recognition or a better understanding of the causal pathways.
### 2) Framework of Precision Medicine Process
-The framework of precision medicine is a process made up of number of feedback loops, with no steady-end point. Each cycle is an attempt to make the process more precise with more stratification of patients. The data assessed in the patients are used to try to develop clinically relevant models, and results of these analyses then inform the further assessment of patients as an evolving result.
+The framework of precision medicine is a process made up of a number of feedback loops, with no steady-end point. Each cycle is an attempt to make the process more precise with more stratification of patients. The data assessed in the patients are used to try to develop clinically relevant models, and the results of these analyses then inform the further assessment of patients as an evolving result.

-###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from track 1-3 are fed back to deep phenotyping stage for subsequent assessments.
+###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after a number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from tracks 1-3 are fed back to the deep phenotyping stage for subsequent assessments.
## 2. Deep Phenotyping
### 1) Importance of Phenotype
-The word phenotype can be used differently in biology versus medicine. In biology, ‘phenotype’ is collection of observable physical properties of an organism, including the organism's appearance, development, behavior, and even characteristics such as gene expression profiles in response to environmental cues. In medical context, however, ‘phenotype’ is deviation from normal morphology, physiology, or behavior. When physician makes note of patient’s phenotype, the physician is doing so by taking medical history, performing physical examination, diagnostic imaging, blood tests, and other psychological tests.
+The word phenotype can be used differently in biology versus medicine. In biology, ‘phenotype’ is a collection of observable physical properties of an organism, including the organism's appearance, development, behavior, and even characteristics such as gene expression profiles in response to environmental cues. In a medical context, however, ‘phenotype’ is a deviation from normal morphology, physiology, or behavior. When a physician makes note of a patient’s phenotype, the physician is doing so by taking a medical history, performing a physical examination, diagnostic imaging, blood tests, and other psychological tests.
-This is where precise phenotype information comes into play. As physician makes note of patient’s phenotypes, the physician is making a diagnosis for the patient’s disease – making a hypothesis of what it may or may not be and provides treatment which may or may not work. Making a diagnosis, therefore, is perhaps the most important task in treating a disease, but it is the most challenging task especially for the cases of rare diseases of which 8000 diseases are yet to be classified. Today, major clinical problems arise from delayed or inaccurate diagnosis, treatment and unnecessary procedures that result in patients’ psychological burdens and unnecessary expenses. To prevent complications and set forth effective treatments from making correct prognosis, full spectrum of phenotypic abnormalities is absolutely critical for improving care quality while reducing the need for unnecessary diagnostic testing and therapies. Figure 3 shows importance of phenotype and its molecular network underpinning. Even with the same endophenotype, risk varies between individuals, because their phenotype ultimately varies in the molecular level.
+This is where precise phenotype information comes into play. As the physician makes note of the patient’s phenotypes, the physician is making a diagnosis for the patient’s disease – making a hypothesis of what it may or may not be and provides treatment that may or may not work. Making a diagnosis, therefore, is perhaps the most important task in treating a disease, but it is the most challenging task especially for the cases of rare diseases of which 8000 diseases are yet to be classified. Today, major clinical problems arise from delayed or inaccurate diagnosis, treatment, and unnecessary procedures that result in patients’ psychological burdens and unnecessary expenses. To prevent complications and set forth effective treatments from making correct prognosis, a full spectrum of phenotypic abnormalities is absolutely critical for improving care quality while reducing the need for unnecessary diagnostic testing and therapies. Figure 3 shows the importance of phenotype and its molecular network underpinning. Even with the same endophenotype, risk varies between individuals, because their phenotype ultimately varies at the molecular level.

@@ -51,19 +51,19 @@ This is where precise phenotype information comes into play. As physician makes
### 2) Why "Deep" Phenotyping?
-The major problem with current phenotyping regime is sloppy or imprecise descriptions of phenotypic descriptions in medical publications. For instance, “still walking after 25 years of onset” is definitely a phenotypic description, yet it is unclear what the physician observed in the patient as such descriptions are likely to evoke different ideas for different readers depending on personal experience and imagination. Furthermore, the current gene mutation database has very little to no phenotypic information. Other than the fact that the patient has that mutagen. While this information is indeed important for determining pathogenicity, it is devoid of natural history of the disease – which is very important clinically in understanding the spectrum of complications with the disease, or genotype-phenotype correlations. Deep phenotyping offers solutions for current challenges, including semantics and technical standards for phenotype and disease data. Deep phenotyping is defined as precise and comprehensive analysis of phenotypic abnormalities in which individual components of individual components of the phenotype are observed and described. Figure 4 shows precision medicine approach to phenotyping; even the individuals with same endophenotype may are biologically distinct and encompass different disease profiles.
+The major problem with the current phenotyping regime is sloppy or imprecise descriptions of phenotypic descriptions in medical publications. For instance, “still walking after 25 years of onset” is definitely a phenotypic description, yet it is unclear what the physician observed in the patient as such descriptions are likely to evoke different ideas for different readers depending on personal experience and imagination. Furthermore, the current gene mutation database has very little to no phenotypic information. Other than the fact that the patient has that mutagen. While this information is indeed important for determining pathogenicity, it is devoid of the natural history of the disease – which is very important clinically in understanding the spectrum of complications with the disease, or genotype-phenotype correlations. Deep phenotyping offers solutions for current challenges, including semantics and technical standards for phenotype and disease data. Deep phenotyping is defined as a precise and comprehensive analysis of phenotypic abnormalities in which individual components of the phenotype are observed and described. Figure 4 shows a precision medicine approach to phenotyping; even the individuals with the same endophenotype may are biologically distinct and encompass different disease profiles.

-##### Figure 4: Individuals undergo deep phenotyping process with data analysis performed using network analysis. This is deep phenotyping, because individuals are classified beyond their endophenotypes. Such methodology optimizes identification of biomarkers, clinical trials, and prognostication of disease.
+##### Figure 4: Individuals undergo a deep phenotyping process with data analysis performed using network analysis. This is deep phenotyping because individuals are classified beyond their endophenotypes. Such methodology optimizes the identification of biomarkers, clinical trials, and prognostication of disease.
### 3) Processing Deep Phenotyping Data
-Conversion of deep phenotyping data into tangible therapeutic utility poses number of challenges yet to be solved. First, computational environments for analysis of high dimensional data is required. Second, there must be data from large populations of patients as a knowledge network. The most important, however, is the integration of the two. For example, next-Generation sequencing enables genome-wide investigation of rare genetic variants, but associating these with diseases requires tailored statical tools that summarize the information of neighbored variants. Indeed, the integration of two types of different information are critical in generating the data in clinically relevant way. To clarify this, the structure of the precision medicine process offers tracks 1-3, which are handling of the data for better prediction of drug sensitivity.
+Conversion of deep phenotyping data into tangible therapeutic utility poses a number of challenges yet to be solved. First, computational environments for analysis of high dimensional data is required. Second, there must be data from large populations of patients as a knowledge network. The most important, however, is the integration of the two. For example, next-Generation sequencing enables genome-wide investigation of rare genetic variants, but associating these with diseases requires tailored statical tools that summarize the information of neighbored variants. Indeed, the integration of two types of different information are critical in generating the data in a clinically relevant way. To clarify this, the structure of the precision medicine process offers tracks 1-3, which are handling of the data for better prediction of drug sensitivity.
## 3. Data Analysis
-Large-scale of data is expected to be integrated and converted into more precise therapeutic interventions. The analysis of deep phenotyping is distinguished into three sequential tracks: in track 1, the data are handled without knowledge of a clinical end-point; in track 2, data are used to build models for a more precise diagnosis or prognosis of disease or disease outcome; and track 3 leads to models that predict more precisely how well specific patients respond to treatment. [Afzal]
+Large-scale data is expected to be integrated and converted into more precise therapeutic interventions. The analysis of deep phenotyping is distinguished into three sequential tracks: in track 1, the data are handled without knowledge of a clinical end-point; in track 2, data are used to build models for a more precise diagnosis or prognosis of disease or disease outcome; and track 3 leads to models that predict more precisely how well specific patients respond to treatment. [Afzal]

@@ -71,7 +71,7 @@ Large-scale of data is expected to be integrated and converted into more precise
### 1) Preprocessing and Data Mining
-The first step of data analysis is always data preprocessing. Studies on fractional exhaled nitric oxide as a marker of eosinophilic inflammation in patients with asthma might be influenced by variable techniques of measurement, sampling procedures, breathing manoeuvres and different types of devices. Therefore, data is expected to be filtered and adjusted to be comparable in analysis. **Quality Control** and **Preprocessing** of the data shall be preformed in this step. Since quality control is always dependent on the specific data type, more details can be found in paper. [Ferté ]
+The first step of data analysis is always data preprocessing. Studies on fractional exhaled nitric oxide as a marker of eosinophilic inflammation in patients with asthma might be influenced by variable techniques of measurement, sampling procedures, breathing maneuvers, and different types of devices. Therefore, data is expected to be filtered and adjusted to be comparable in the analysis. **Quality Control** and **Preprocessing** of the data shall be performed in this step. Since quality control is always dependent on the specific data type, more details can be found in the paper. [Ferté ]
Some of the common approaches for data preprocessing includes:
@@ -89,7 +89,7 @@ Some of the common approaches for data preprocessing includes:
**Case Example:** *Denoising + Variable selection*
-*The collection of deep phenotyping data usually includes great amount of unselected data, which will inevitably include variables that are irrelevant for later modelling. The irrelevant variables that contain random “noise” can mask important underlying relationships and structures, which can shown in below.*
+*The collection of deep phenotyping data usually includes a great amount of unselected data, which will inevitably include variables that are irrelevant for later modeling. The irrelevant variables that contain random “noise” can mask important underlying relationships and structures, which is shown below.*
45 individuals are clustered into three homogeneous subgroups if two variables are considered. A hierarchical cluster analysis using only these two variables identifies three clusters clearly (figure 6b). However, if three random noise variables are added to the dataset, the cluster algorithm fails to find the three groups as indicated in figure 6c. We can see that about half of the individuals are classified incorrectly.
@@ -97,11 +97,11 @@ Some of the common approaches for data preprocessing includes:
##### Figure 6: Impact of variable selection based on illustration
-Variable selection is important for filtering out noise data, yet how to define noise or whether a variable is revevant is always a unanswerable question. One important factor that can guide variable selection is scientifically based *a priori* knowledge about the possibly **biologically relevant variables**.
+Variable selection is important for filtering out noise data, yet how to define noise or whether a variable is relevant is always an unanswerable question. One important factor that can guide variable selection is scientifically based a priori knowledge about the possibly **biologically relevant variables**.
### 2) Diagnostic and Prognostic Models
-After data is ready from track 1 for further training, models can be developed to estimate the risk with absolute probability of the presence or absence of an outcome or disease in individuals based on their clinical and non-clinical information. Prediction model can be diagnostic (outcome or disease present at this moment) or prognostic (outcome occurs within a specified time frame). [Hendriksen]
+After data is ready from track 1 for further training, models can be developed to estimate the risk with an absolute probability of the presence or absence of an outcome or disease in individuals based on their clinical and non-clinical information. Prediction model can be diagnostic (outcome or disease present at this moment) or prognostic (outcome occurs within a specified time frame). [Hendriksen]

@@ -110,16 +110,16 @@ After data is ready from track 1 for further training, models can be developed t
- Model Development
Two main strategies are usually used to develop the models:
1. **Full Model**
- No predictor selection is applied . Pros: avoid improper predictor selection due to predictor selection bias. Cons: requires much prior knowledge to adequately preselect the biologically relevant predictors for modelling.
+ No predictor selection is applied. Pros: avoid improper predictor selection due to predictor selection bias. Cons: requires much prior knowledge to adequately preselect the biologically relevant predictors for modeling.
2. **Variable selection**
- **Backward Elimination** of ‘redundant’ predictors or **Forward Selection** of ‘promising’ ones. The backward procedure is initialized with the full multivariable and then subsequently removes predictors based on a predefined criterion. Forward selection is when predictors are added to the multivariable model one by one. There is no agreement on the optimal method among the two methods, The choice of methods is highly *Content-Specific*.
+ **Backward Elimination** of ‘redundant’ predictors or **Forward Selection** of ‘promising’ ones. The backward procedure is initialized with the full multivariable and then subsequently removes predictors based on a predefined criterion. Forward selection is when predictors are added to the multivariable model one by one. There is no agreement on the optimal method among the two methods, The choice of methods is highly content-specific.
- Outcome
-The outcome of a prediction aims to reflects a clinically significant and patient relevant health state, for example, death yes or no, or absence or presence of Leukemia. In case of prognostic prediction model, a follow-up period is needed to be clearly defined for outcome development. [Hendriksen]
+The outcome of a prediction aims to reflect a clinically significant and patient-relevant health state, for example, death yes or no, or absence or presence of Leukemia. In the case of the prognostic prediction model, a follow-up period is needed to be clearly defined for outcome development. [Hendriksen]
- Clinics-friendly Accommodation
-Regression model can be too complicated to use in daily clinical uses, the model is usually simplify by rounding coefficient toward integer numbers for easier scoring. However, such accommodation might negatively effected the accuracy of the model and thus needs to be applied carefully.
+The regression model can be too complicated to use in daily clinical uses, the model is usually simplified by rounding coefficient toward integer numbers for easier scoring. However, such accommodation might negatively affected the accuracy of the model and thus needs to be applied carefully.
### 3) Predicting Treatment Response
@@ -127,13 +127,13 @@ After developing diagnostic and prognostic models from the previous step, there
- Prediction models can be built on the diagnostic and prognostic models from the previous stage as prognostic factors may act as the natural variables to consider when developing prediction models. For instance, epidermal growth factor receptor tyrosine kinase status acts as a prognostic factor for survival in patients with non-small cell lung cancer and a predictive factor for response to the tyrosine kinase inhibitor gefitinib as first-line treatment at the same time [Riley RD(will change later after sorting&numbering sources].
-- The data can be directly utilized to extract significant information relevant to the prediction of the disease. For instance, in the first clinical trials that used the monoclonal anti-IL-5 antibody mepolizumab for asthma treatment, the use of mepolizumab was associated with a significant reduction blood and sputum eosinophils but did not have significant clinical benefit in asthma patients [Flood-Page]. Consequently, in the following trials, patients with refractory eosinophilic asthma were selected and in this subgroup, mepolizumab therapy added significant clinical benefit in patients by reducing exacerbations and improving asthma quality of life scores [Haldar P].
+- The data can be directly utilized to extract significant information relevant to the prediction of the disease. For instance, in the first clinical trials that used the monoclonal anti-IL-5 antibody mepolizumab for asthma treatment, the use of mepolizumab was associated with a significant reduction in blood and sputum eosinophils but did not have significant clinical benefit in asthma patients [Flood-Page]. Consequently, in the following trials, patients with refractory eosinophilic asthma were selected and in this subgroup, mepolizumab therapy added significant clinical benefit in patients by reducing exacerbations and improving asthma quality of life scores [Haldar P].

##### Figure 8: The figure above is an illustration of applying different treatments, also known as "tailored medicine", to categorized groups to increase the effectiveness of treatment based on a prediction model.
-This process of building prediction models involves generating further knowledge about the disease and treatment from the diagnotics and prognostics models or the data itself. The findings from this process, in the form of data, would be fed back to the phenotyping of patients to further adjust the clinical trial that could more precisely show the effectiveness of the treatment.
+This process of building prediction models involves generating further knowledge about the disease and treatment from the diagnostics and prognostics models or the data itself. The findings from this process, in the form of data, would be fed back to the phenotyping of patients to further adjust the clinical trial that could more precisely show the effectiveness of the treatment.
In addition to the feedback, dissemination and communication of the taxonomy and models with the clinical and scientific communities, for instance, to provide utilizable algorithms for clinical practices.
@@ -143,7 +143,7 @@ In addition to the feedback, dissemination and communication of the taxonomy and

-As seen in the figure, the cycle of patients assessment(deep phenotyping), data processing(preprocessing and data mining), and model building(diagnotic, prognostic, and prediction model building) is repeated at least several times to further increase the preciseness of the medicine by categorizing patient groups at a higher resolution.
+As seen in the figure, the cycle of patients assessment(deep phenotyping), data processing(preprocessing and data mining), and model building(diagnostic, prognostic, and prediction model building) is repeated at least several times to further increase the preciseness of the medicine by categorizing patient groups at a higher resolution.
The detailed steps are as follows:
@@ -155,7 +155,7 @@ The detailed steps are as follows:
## 5. Conclusion
-In this paper, the process of precision medicine, starting from the deep phenotyping stage to data analysis part, which comprised preprocessing, data mining, developing diagnostic, prognostic, and prediction models, were discussed against the background of airway diseases. As seen in the previous section on the continual evolution of precision medicine with repeated cycle of the above steps, precision medicine should be viewed as the continuous process of feedback loops rather than a steady-state with an end-point or a specific output from the research. In this context, tailored or stratified medicine resulting from new stratifications can be considered as a makeshift product of the process that will generate new data for the next cycle, thus further increasing the precision.
+In this paper, the process of precision medicine, starting from the deep phenotyping stage to the data analysis part, which comprised preprocessing, data mining, developing diagnostic, prognostic, and prediction models, was discussed against the background of airway diseases. As seen in the previous section on the continual evolution of precision medicine with the repeated cycles of the above steps, precision medicine should be viewed as the continuous process of feedback loops rather than a steady-state with an end-point or a specific output from the research. In this context, tailored or stratified medicine resulting from new stratifications can be considered as a makeshift product of the process that will generate new data for the next cycle, thus further increasing the precision.
Once again, the ultimate goal of precision medicine is the most effective treatment for the individual patient. To step closer to this goal, ongoing endeavors to be even more precise and individualistic, newly-gained scientific and clinical knowledge, and new data sources would be necessary.
From f8953f38a073e9cf81753ff8d820974442849556 Mon Sep 17 00:00:00 2001
From: Soyeon Kim <37919264+ksoyeona@users.noreply.github.com>
Date: Wed, 16 Dec 2020 09:59:46 +0900
Subject: [PATCH 64/94] Update Precision_Medicine.md
---
.../Precision_Medicine.md | 30 ++++++++++---------
1 file changed, 16 insertions(+), 14 deletions(-)
diff --git a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
index 1aae017..f4abb61 100644
--- a/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
+++ b/finalPaper/Grup25_Precision_Medicine/Precision_Medicine.md
@@ -21,13 +21,13 @@
### 1) Definition
-Precision medicine is personalized treatment strategies on the basis of genetic, biomarker, phenotypic or psychosocial characteristics to stratify patients into novel subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. It is also commonly referred to as 'stratified medicine', 'targeted therapy', and 'deep phenotyping.'
+Precision medicine is personalized treatment strategies on the basis of genetic, biomarker, phenotypic or psychosocial characteristics to stratify patients into novel subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. It is also commonly referred to as 'stratified medicine', 'targeted therapy', and 'deep phenotyping'.
The focus of applied definitions is moving away from the classical 'signs-and-symptoms’ approach or 'population’ approach into the 'N of 1’ approach in which each patient is an entire trial as a single case study from N number of individuals within a population. Figure 1 shows the difference between two approaches with the same group of patients. A comprehensive study of such subclasses ultimately depends on computational resources to capture, store and exchange phenotypic data, and upon sophisticated algorithms to integrate it with genomic variation, omics profiles, and other clinical information.

-##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and give the same treatment, because they were treated as one group for the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group is divided into disease subgroups with more precise and validated phenotypic recognition or a better understanding of the causal pathways.
+##### Figure 1: In the left diagram, colon cancer patients are clustered into one group and are given the same treatment, because they are treated as one group for the same symptoms. This does not take into account that patients may respond differently to a particular disease or treatment. On the right diagram, however, the same group is divided into disease subgroups with more precise and validated phenotypic recognition or a better understanding of the causal pathways.
### 2) Framework of Precision Medicine Process
@@ -35,7 +35,7 @@ The framework of precision medicine is a process made up of a number of feedback

-###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoise and baseline correction and variable selection. Track 2 forecasts clinically relevant outcome as the model after a number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from tracks 1-3 are fed back to the deep phenotyping stage for subsequent assessments.
+###### Figure 2: This diagram shows the process of precision medicine. In the deep phenotyping stage, data are gathered and forwarded to tracks 1-3, which are clinical data assessment steps. In track 1, data are preprocessed and explored for previously unknown structure with data mining techniques, such as clustering analysis to derive subgroup of samples. This step also composes of denoising, baseline correction, and variable selection. Track 2 forecasts clinically relevant outcome as the model after a number of feedback loops. Track 3 predicts treatment response and generates further knowledge about the treatment. This step also utilizes feedback to patient phenotyping results. Results from tracks 1-3 are fed back to the deep phenotyping stage for subsequent assessments.
## 2. Deep Phenotyping
@@ -51,7 +51,7 @@ This is where precise phenotype information comes into play. As the physician ma
### 2) Why "Deep" Phenotyping?
-The major problem with the current phenotyping regime is sloppy or imprecise descriptions of phenotypic descriptions in medical publications. For instance, “still walking after 25 years of onset” is definitely a phenotypic description, yet it is unclear what the physician observed in the patient as such descriptions are likely to evoke different ideas for different readers depending on personal experience and imagination. Furthermore, the current gene mutation database has very little to no phenotypic information. Other than the fact that the patient has that mutagen. While this information is indeed important for determining pathogenicity, it is devoid of the natural history of the disease – which is very important clinically in understanding the spectrum of complications with the disease, or genotype-phenotype correlations. Deep phenotyping offers solutions for current challenges, including semantics and technical standards for phenotype and disease data. Deep phenotyping is defined as a precise and comprehensive analysis of phenotypic abnormalities in which individual components of the phenotype are observed and described. Figure 4 shows a precision medicine approach to phenotyping; even the individuals with the same endophenotype may are biologically distinct and encompass different disease profiles.
+The major problem with the current phenotyping regime is unsystematic or imprecise descriptions of phenotypic descriptions in medical publications. For instance, “still walking after 25 years of onset” is definitely a phenotypic description, yet it is unclear what the physician observed in the patient as such descriptions are likely to evoke different ideas for different readers depending on personal experience and imagination. Furthermore, the current gene mutation database has very little to no phenotypic information other than the fact that the patient has that mutagen. While this information is indeed important for determining pathogenicity, it is devoid of the natural history of the disease – which is very important clinically in understanding the spectrum of complications with the disease, or genotype-phenotype correlations. Deep phenotyping offers solutions for current challenges, including semantics and technical standards for phenotype and disease data. Deep phenotyping is defined as a precise and comprehensive analysis of phenotypic abnormalities in which individual components of the phenotype are observed and described. Figure 4 shows a precision medicine approach to phenotyping; even the individuals with the same endophenotype may are biologically distinct and encompass different disease profiles.

@@ -59,11 +59,11 @@ The major problem with the current phenotyping regime is sloppy or imprecise des
### 3) Processing Deep Phenotyping Data
-Conversion of deep phenotyping data into tangible therapeutic utility poses a number of challenges yet to be solved. First, computational environments for analysis of high dimensional data is required. Second, there must be data from large populations of patients as a knowledge network. The most important, however, is the integration of the two. For example, next-Generation sequencing enables genome-wide investigation of rare genetic variants, but associating these with diseases requires tailored statical tools that summarize the information of neighbored variants. Indeed, the integration of two types of different information are critical in generating the data in a clinically relevant way. To clarify this, the structure of the precision medicine process offers tracks 1-3, which are handling of the data for better prediction of drug sensitivity.
+Conversion of deep phenotyping data into tangible therapeutic utility poses a number of challenges yet to be solved. First, computational environments for analysis of high dimensional data is required. Second, there must be data from large populations of patients as a knowledge network. The most important, however, is the integration of the two. For example, next-generation sequencing enables genome-wide investigation of rare genetic variants, but associating these with diseases requires tailored statical tools that summarize the information of neighbored variants. Indeed, the integration of two types of different information are critical in generating the data in a clinically relevant way. To clarify this, the structure of the precision medicine process offers tracks 1-3, which are handling of the data for better prediction of drug sensitivity.
## 3. Data Analysis
-Large-scale data is expected to be integrated and converted into more precise therapeutic interventions. The analysis of deep phenotyping is distinguished into three sequential tracks: in track 1, the data are handled without knowledge of a clinical end-point; in track 2, data are used to build models for a more precise diagnosis or prognosis of disease or disease outcome; and track 3 leads to models that predict more precisely how well specific patients respond to treatment. [Afzal]
+Large-scale data is integrated and converted into more precise therapeutic interventions. The analysis of deep phenotyping is distinguished into three sequential tracks: in track 1, the data are handled without knowledge of a clinical end-point; in track 2, data are used to build models for a more precise diagnosis or prognosis of disease or disease outcome; and track 3 leads to models that predict more precisely how well specific patients respond to treatment. [Afzal]

@@ -71,7 +71,7 @@ Large-scale data is expected to be integrated and converted into more precise th
### 1) Preprocessing and Data Mining
-The first step of data analysis is always data preprocessing. Studies on fractional exhaled nitric oxide as a marker of eosinophilic inflammation in patients with asthma might be influenced by variable techniques of measurement, sampling procedures, breathing maneuvers, and different types of devices. Therefore, data is expected to be filtered and adjusted to be comparable in the analysis. **Quality Control** and **Preprocessing** of the data shall be performed in this step. Since quality control is always dependent on the specific data type, more details can be found in the paper. [Ferté ]
+The first step of data analysis is data preprocessing. Studies on fractional exhaled nitric oxide as a marker of eosinophilic inflammation in patients with asthma might be influenced by variable techniques of measurement, sampling procedures, breathing maneuvers, and different types of devices. Therefore, data is filtered and adjusted to be comparable in the analysis. **Quality Control** and **preprocessing** of the data are performed in this step [Ferté ].
Some of the common approaches for data preprocessing includes:
@@ -89,15 +89,15 @@ Some of the common approaches for data preprocessing includes:
**Case Example:** *Denoising + Variable selection*
-*The collection of deep phenotyping data usually includes a great amount of unselected data, which will inevitably include variables that are irrelevant for later modeling. The irrelevant variables that contain random “noise” can mask important underlying relationships and structures, which is shown below.*
+*The collection of deep phenotyping data includes a great amount of unselected data, which will inevitably include variables that are irrelevant for later modeling. The irrelevant variables that contain random “noise” can mask important underlying relationships and structures, which is shown below.*
-45 individuals are clustered into three homogeneous subgroups if two variables are considered. A hierarchical cluster analysis using only these two variables identifies three clusters clearly (figure 6b). However, if three random noise variables are added to the dataset, the cluster algorithm fails to find the three groups as indicated in figure 6c. We can see that about half of the individuals are classified incorrectly.
+In the figure below, 45 individuals are clustered into three homogeneous subgroups if two variables are considered. A hierarchical cluster analysis using only these two variables identifies three clusters clearly in figure 6b. However, if three random noise variables are added to the dataset, the cluster algorithm fails to find the three groups as indicated in figure 6c. About half of the individuals are classified incorrectly.

##### Figure 6: Impact of variable selection based on illustration
-Variable selection is important for filtering out noise data, yet how to define noise or whether a variable is relevant is always an unanswerable question. One important factor that can guide variable selection is scientifically based a priori knowledge about the possibly **biologically relevant variables**.
+Variable selection is important for filtering out noisy data, yet how to define noise or whether a variable is relevant remains unanswerable. One important factor that can guide variable selection is scientifically based a priori knowledge about the possibly **biologically relevant variables**.
### 2) Diagnostic and Prognostic Models
@@ -110,16 +110,18 @@ After data is ready from track 1 for further training, models can be developed t
- Model Development
Two main strategies are usually used to develop the models:
1. **Full Model**
- No predictor selection is applied. Pros: avoid improper predictor selection due to predictor selection bias. Cons: requires much prior knowledge to adequately preselect the biologically relevant predictors for modeling.
+ No predictor selection is applied.
+ Pros: avoid improper predictor selection due to predictor selection bias.
+ Cons: requires much prior knowledge to adequately preselect the biologically relevant predictors for modeling.
2. **Variable selection**
**Backward Elimination** of ‘redundant’ predictors or **Forward Selection** of ‘promising’ ones. The backward procedure is initialized with the full multivariable and then subsequently removes predictors based on a predefined criterion. Forward selection is when predictors are added to the multivariable model one by one. There is no agreement on the optimal method among the two methods, The choice of methods is highly content-specific.
- Outcome
-The outcome of a prediction aims to reflect a clinically significant and patient-relevant health state, for example, death yes or no, or absence or presence of Leukemia. In the case of the prognostic prediction model, a follow-up period is needed to be clearly defined for outcome development. [Hendriksen]
+The outcome of a prediction aims to reflect a clinically significant and patient-relevant health state, for example, prediction on death, or absence or presence of leukemia. In the case of the prognostic prediction model, a follow-up period is needed to be clearly defined for outcome development. [Hendriksen]
- Clinics-friendly Accommodation
-The regression model can be too complicated to use in daily clinical uses, the model is usually simplified by rounding coefficient toward integer numbers for easier scoring. However, such accommodation might negatively affected the accuracy of the model and thus needs to be applied carefully.
+The regression model can be too complicated to use in daily clinical uses. Therefore, the model is usually simplified by rounding coefficient toward integer numbers for easier scoring. However, such accommodation might negatively affect the accuracy of the model and thus it needs to be applied carefully.
### 3) Predicting Treatment Response
@@ -177,7 +179,7 @@ Flood-Page P, Swenson C, Faiferman I, et al. A study to evaluate safety and effi
Haldar P, Brightling CE, Hargadon B, et al. Mepolizumab and exacerbations of refractory eosinophilic asthma. N Engl J Med 2009; 360: 973–984.
-the image- might not use-- “What Is Precision Medicine and How Can It Help Fix Healthcare.” ReferralMD, ReferralMD, 11 Dec. 2018, getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/. https://getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/
+“What Is Precision Medicine and How Can It Help Fix Healthcare.” ReferralMD, ReferralMD, 11 Dec. 2018, getreferralmd.com/2018/02/precision-medicine-can-help-fix-healthcare/.
Ferté C, Trister AD, Huang E, et al. Impact of bioinformatic procedures in the development and translation of high-throughput molecular classifiers in oncology. Clin Cancer Res 2013; **19**: 4315–4325.
From dd339a5b16d178c5883b8a3d318d3acdbefe308b Mon Sep 17 00:00:00 2001
From: meixwu v{dW#L
z$pjv}Twdpqrsw>;d6yRo6)mezKx6P3)l(Q#!e**{z&n#oN{zmteG*OQ@ZM~nm0n)O;-gaG*lJ@LW~w9#
z)%eonc7#5l@39sCL}knon=#~ZuW&gBFXe8?*v~S$Y^s7~bd|5W6vIIABoZz2W-!&>
z{|!Fan|+XA#IZHvf8La(rNnu{oFnb79g3zn)HhzDfCfB}W=+HoPB05o{F3VMlI32Q
zloU?
rE5NH^Lc
zS1xt&V0!Ik^+4f+sIU1qiSN;0OiL{j<3h-*T`o?QeU^QsG`(GDoe>D%F%U-;q8v&v*;7L{9BKwC(bu
z4;UuUk#D!pp97^V6NDvr-`N+kI>>UR+FvBMugOvPDn4O@`Sv8cXH!k}a}*7VmSc4F
zZ7*KIPk%PgVtYIodH@Gkdh}5?Az@d;FTgtD-z%-(7i_vxe(9e}mC`_cd2}_^lDm8u
z(-?H^RI2S1`7^df9hJ_%)WZp+f72^~AtN^+
jB4y8Ux+KQ)ZUL8=~;z6yU%s-9(5>6GoU>WNS9|z|{UE