Machine learning, network-based models for gene expression, activity, function

Division Of Basic Sciences - Nci

Investigators

Linked publications & trials

Abstract

In this analysis, we utilize GTex and TCGA for bulk data, Human Cell Landscape for scRNA mRNA, and GSE81287 scRNA miRNA data. First, we established that a model based on all genes performs better than one based only on annotated targets of a miRNA. When we applied our method to bulk data, we observed that our method works well between normal and cancer tissues - average correlation coefficient between predicted and actual miRNA levels is 0.5. In general the model works better when larger heterogeneous data are used for training. We are currently working on the validation part. A manuscript reporting this work about to be sent out for review. We have applied PathExt tool to SARS-CoV-2 response data in cell lines and in patient PBMC and identify several novel genes and suggest novel drugs against COVID-19. We are currently preparing the manuscript reporting this. We are also starting to apply the tool to Breast tumor transcriptomic data to identify central genes driving various subtypes of breast cancer.