Discovery of novel therapeutic targets in cancer using patient-specific gene regulatory networks

Abstract

Most cancer types lack effective targeted therapeutic options and in cancers where first-line targeted therapies are available, treatment resistance is a huge challenge. Recent technological advances enable the use of ATAC-seq and RNA-seq on patient biopsies in a high-throughput manner. Here we present a computational approach that leverages these datasets to identify novel drug targets based on tumor lineage. We constructed patient-specific gene regulatory networks for 371 patients of 22 cancer types using machine learning approaches trained using three-dimensional genomic data for enhancer to promoter contacts. Next, we identify the key transcription factors (TFs) in these networks, which are used to identify therapeutic vulnerabilities either by direct targeting of TFs or proteins that they co-operate with. We validate four novel candidates identified for neuroendocrine, liver and renal cancers, which have a dismal prognosis with current therapeutic options. We present a novel approach to use the increasing amounts of functional genomics data from patient biospecimens for identification of novel drug targets.

Recapitulation of patient-specific 3D chromatin conformation using machine learning

“… 52 , 54 Furthermore, the identified enhancer-gene connections make it possible for constructing more precise TF to gene networks for the 371 patients with primary cancer to facilitate the discovery of important TFs and potential novel drug targets. 55 …”

Dozer: Debiased personalized gene co-expression networks for population-scale scRNA-seq data

“…Next, we considered the impact of expression differences due to batch effects on gene centrality estimates from the co-expression networks. With population scale data, a key downstream analysis is to detect network-level differences associated with phenotypic or genotypic variation (Forbes, 2022). If a network construction method is biased by the actual expression levels of the genes, differential expression will impact the detection of network-level changes.…”

“…A key opportunity unveiled by emerging scRNA-seq datasets is the construction of personalized gene co-expression networks which can be leveraged to link network-level properties to phenotypic variation, e.g., discovering therapeutic targets in cancer (Forbes, 2022) and identifying genetic variants (e.g., network QTLs) that associate with network properties such as modules (Langfelder and Horvath, 2008) and network centrality (Savino et al, 2020) measures. Gene co-expression network analysis (Zhang and Horvath, 2005), which estimates gene-gene correlations, is a key inference tool for detecting latent relationships that might be obscured in standard analysis of clustering and differential expression.…”

Drug repurposing for viral cancers: A paradigm of machine learning, deep learning, and virtual screening‐based approaches

“…The notable databases include The Cancer Genome Atlas (TCGA), 34 Gene Expression Omnibus, 35,36 and Genotype-Tissue Expression [37][38][39] etc. These databases have collected, analyzed, and curated the huge amount of patient-derived and patient-specific data which have led to the identification of novel therapeutic targets [40][41][42] as well as the analysis of well-known targets in required disease areas such as viral cancers. 43 Computational methods to find the repurposable drugs targeting the viral cancers use supervised ML and DL methods and make use of the publicly accessible databases 28,[44][45][46][47] and in-house information resources.…”