PacEnClust
PacEnClust (an unsupervised ensemble clustering model for Identifying Pancreatic Cancer Subtypes), is an effective wMetaC-based method for Pancreatic Cancer subtype prediction based on transcriptomic data. Leveraging multiple runs over different randomly projected sets of data, PacEnClust is able to seperate a 1048 sample dataset into five different Pancreatic Cancer subtypes, which may provide insights into the characteristics of these subtypes that can significantly aid clinical decision-making processes.
- Download the PacEnClust code along with the dataset (WmetaCXzc.rds) and labels (batches_for_mats.txt) from the github page
- Specify the current directory in all sections containing the phrase "replace with target directory", following the hashtagged directions
- Put WmetaCXzc.rds (test data) or your own Z-scored input data (must be structured as below, with samples as columns and genes as rows) in this directory
- run the entire block of code, the cluster labels will be saved as "TCGAexpressionCLUSTER.csv" and the evaluation metric scores will be seen in the output as a dataframe named "metrics_df"
- specify a manually chosen number of clusters or automatically find the optimal number based on silhouette score
- chose by running either the first version of the getrowcolor() function, which will automatically choose the cluster amount (k) or running the second, in which a target cluster amount (k) can be specified
If you find any bugs or problems, or you have any comments, please don't hesitate to contact nickpeterson@unmc.edu via email or post to Issues.
Nick Peterson, Jieqiong Wang, Shibiao Wan
PacEnClust for Pancreatic Cancer Subtype Prediction Nick Peterson, Jieqiong Wang, Shibiao Wan bioRxiv TBD; doi: TBD
GNU GENERAL PUBLIC LICENSE
Version 3, 29 June 2007
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.