Skip to main content
Premium Trial:

Request an Annual Quote

Systems Study Taps Cancer Gene Expression Reversal to Predict Drug Response

NEW YORK (GenomeWeb) – Drugs that prompt promising anti-cancer responses in preclinical disease models also lead to gene expression shifts in related cell lines that may be useful for predicting drug effectiveness, new research suggests.

"Using gene expression as a representation of the molecular state, a number of studies have demonstrated its potential in drug discovery; yet, there was no systematic way to correlate reversal potency and drug efficacy," senior author Atul Butte, a researcher at the University of California at San Francisco, and his co-authors wrote in their new Nature Communications study.

"Our study leveraged the emerging public cancer genomics and pharmacogenomics databases to address this challenge," they explained, "and we successfully demonstrated that reversal potency correlates with drug efficacy and can be used to predict potential new drug candidates for several cancer types."

As they reported online today, the researchers established cancer-related gene expression signatures based on RNA sequence data for thousands of samples from 14 cancer types that were assessed for the Cancer Genome Atlas project. They considered these signatures in relation to the expression patterns produced when dozens of cell lines were exposed to more than 12,400 drug compounds, uncovering expression changes that seemed to accompany drug response.

"[W]e show that the potency of a drug to reverse cancer-associated gene expression changes positively correlates with that drug's efficacy in preclinical models of breast, liver, and colon cancers," the authors wrote.

With the help of a systems-based approach, the team narrowed in on four compounds with predicted activity against liver cancer, in particular. And when the group tested the compounds in five liver cancer cell lines, it saw diminishing liver cancer cell viability as the concentration of these drugs was dialed up.

For the first stage of their analysis, the researchers brought together TCGA RNA sequence data for 689 normal and 6,825 tumor samples spanning 14 cancer types, narrowing in on differentially expressed genes that marked cancer-related expression signatures.

The team considered these signatures in relation to more than 66,600 expression profiles in the Library of Integrated Network-based Cellular Signatures (LINCS) produced across 71 cell lines that were exposed to up to 12,442 compounds.

As part of the process of narrowing in on breast invasive carcinoma, liver hepatocellular carcinoma, and colon adenocarcinoma for further analyses, the researchers folded in molecular profiles for 1,046 cell lines representing 10 cancer types from the Cancer Cell Line Encyclopedia, along with millions of drug activity measurements from the ChEMBL database, which encompassed more than 1.1 million compounds and 1,647 cell lines.

They also used these data to identify drugs suspected of reversing the cancer-related gene expression signature, prompting drug response predictions that were tested in cell lines that coincided with the cancer types considered — experiments supporting the notion that the extent of gene expression reversal was tied to potential drug potency.

The team's work in liver hepatocellular took this a step further, narrowing in on four candidate compounds: pyrvinium pamoate, strophanthidin, FCCP, and CGK 733. Indeed, the group reported that each of the compounds did dial down the viability of cells from five liver hepatocellular cell lines, suggesting the systems-based approach may be capable of picking up promising drug compounds.

Based on their initial findings, they suggested that expression data and broader systems analyses "may be complementary to the traditional target-based approach in connecting diseases to potentially efficacious drugs."