Probabilistic Models for Gene Silencing Data

Abstract

This thesis is concerned with signaling pathways leading to regulation of gene expression.I develop methodology to address two problems specific to gene silencing experiments:First, gene perturbation effects cannot be controlled deterministically and have to bemodeled stochastically. Second, direct observations of intervention effects on other path-way components are often not available.

Cross-talk between AMPK and EGFR dependent Signaling in Non-Small Cell Lung Cancer

“…NEMs (Nested Effects Models) are a class of probabilistic graphical models used to reconstruct a pathway based on measurable downstream response profiles of individual gene knockdowns 15 43 . Gene silencing typically alters steady-state expression/activity levels of downstream transcripts 44 45 .…”

“…The approach for network reconstruction is based on single siRNA based knockdowns of each gene in the H1650 cell line (EGFR, delE746-A750) and subsequent gene expression profiling. Based on these data we employed Nested Effects Models (NEMs) as a statistical learning approach to unravel key elements of the interplay between AMPK and EGFR dependent signaling 15 . The resulting network is then validated using protein expression data in cell lines and lung ADC patient data (RNAseq plus somatic mutations) from The Cancer Genome Atlas (TCGA), demonstrating the relevance of our findings.…”