A Gene Expression High-Throughput Screen (GE-HTS) for Coordinated Detection of Functionally Similar Effectors in Cancer

Abstract

Genome-wide, loss-of-function screening can be used to identify novel vulnerabilities upon which specific tumor cells depend for survival. Functional Signature Ontology (FUSION) is a gene expression-based high-throughput screening (GE-HTS) method that allows researchers to identify functionally similar proteins, small molecules, and microRNA mimics, revealing novel therapeutic targets. FUSION uses cell-based high-throughput screening and computational analysis to match gene expression signatures produced by natural products to those produced by small interfering RNA (siRNA) and synthetic microRNA libraries to identify putative protein targets and mechanisms of action (MoA) for several previously undescribed natural products. We have used FUSION to screen for functional analogues to Kinase suppressor of Ras 1 (KSR1), a scaffold protein downstream of Ras in the Raf-MEK-ERK kinase cascade, and biologically validated several proteins with functional similarity to KSR1. FUSION incorporates bioinformatics analysis that may offer higher resolution of the endpoint readout than other screens which utilize Boolean outputs regarding a single pathway activation (i.e., synthetic lethal and cell proliferation). Challenges associated with FUSION and other high-content genome-wide screens include variation, batch effects, and controlling for potential off-target effects. In this review, we discuss the efficacy of FUSION to identify novel inhibitors and oncogene-induced changes that may be cancer cell-specific as well as several potential pitfalls within FUSION and best practices to avoid them.

Keywords: Ras-driven cancer; cancer susceptibility genes; functional signature ontology; high-throughput screens.

Figure 1

Evaluation of negative control normalization based on precision and scalability: (a) scatterplot of Kinase suppressor of Ras 1 (KSR1)-depleted wells (red dots) and individual gene depletions from the siGenome library (black dots) based on Euclidean distance (arbitrary distance units) and Pearson correlation similarity metrics and (b) scatterplot showing correlation of the Euclidean distance metrics after normalizing to the nontargeting control wells using the ranking of previously validated kinome hits calculated from the kinome only analysis (x-axis) vs. the entire-genome screen analysis (y-axis).