ClinicalTrials_Prediction

This repo contains the source code for "Predicting Clinical Trial Outcomes Using Drug Bioactivities through Graph Database Integration and Ensemble Learning"

Artifacts: Data Integration of multiple heterogenous chemical databases to Neo4j. These databases are DrugBank, ChEMBL, AACT, CTD, and UniProt.

DataFusion: Mapping across database and within a database

Machine learning: Inhouse curated data extraction from the integrated data source, centered around ClinicalTrials, Compounds, Diseases and Targets Feature extraction for ML Model Training, paramter tuning with supervised algorithms and performance measures

Pre-requisites for Neo4j DB Integration:

1. Neo4J-Community edition is installed ( We used 4.1.3 version of Unix distribution inside anaconda environment ).

2. Open neo4j_community/conf/neo4j.conf and change the following lines.

   1. Comment this line "dbms.directories.import=import"
   2. Change the value of this line "dbms.security.auth_enabled=true" to false.
   3. Change the value of this line "dbms.memory.heap.max_size=1G" to 70G
   4. Add this line "dbms.security.allow_csv_import_from_file_urls=true"

Now,to start the graph, open terminal and go to neo4j community folder and run the following command : ./bin/neo4j console

Note:

The integrated data can be download from Figshare : https://doi.org/10.6084/m9.figshare.18631901     
Copy the unzipped file to data folder of Neo4j (e.g. neo4j-community-4.1.3/data/databases)

Citation

If you find the work useful in your research, please consider citing:

Murali, V., Pradyumna, Y. M., Königs, C., Nair, M., Nedungadi, P., Srinivasa, G., & Athri, P. (2022). Predicting Clinical Trial Outcomes Using Drug Bioactivities through Graph Database Integration and Machine Learning. Chemical Biology & Drug Design.