Activity Number:
|
248
- Machine Learning in Science and Industry
|
Type:
|
Contributed
|
Date/Time:
|
Monday, July 29, 2019 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
Abstract #304389
|
|
Title:
|
Using a Network-Based Approach to Identify Gene Signatures That Predict Cancer Survival
|
Author(s):
|
Minya Pu* and Judith Varner and Karen Messer
|
Companies:
|
University of California, San Diego and University of California, San Diego and University of California, San Diego
|
Keywords:
|
Network-based approach;
Network propagation;
Gene signature;
Classifier;
ReactomeFI
|
Abstract:
|
To identify a signature that can predict lung cancer survival, mRNA expression values were compared between tumor tissue resident (TR) and bone marrow derived (BMD) macrophages from mice; Significantly differentially expressed genes were found using limma. Models obtained from LASSO using these genes as candidate predictors worked very poorly. Thus, we sought to use a network-based approach. Network propagation via GeneMANIA or GeNets based on known networks was used to further include the genes that were functionally related to the DE genes. Module detection was then performed to form functional gene clusters, using methods implemented in WGCNA or cyto-ReactomeFI. Final candidate gene list used genes in functional clusters with desired functions and the genes that connect these modules. The PAM method was used to make nearest shrunken centroid classifier. The classifier was validated using corresponding TCGA human data: all the patient tumors were first classified as TR or BMD; the association between the classified subtypes and survival outcomes were then assessed. Future analysis will consider incorporating copy number and mutation data to help prioritizing genes.
|
Authors who are presenting talks have a * after their name.