New study aims to improve predictions of response to cancer drugs for all types of cancer

The goal of “precision oncology” is to be able to tailor treatments to each patient based on the unique molecular fingerprints of their cancer.

New technologies and large “-omics” data sets now allow researchers to examine common characteristics not only within a single cancer type -; like breast cancer -; but to look for patterns in many types of cancer. These data offer excellent clues that an approach that has been shown to be successful in one type of cancer may also work well against a different type of cancer based on common underlying characteristics.

But sifting through this sea of ​​patient data is tricky. A model that predicts that a drug can work against multiple types of cancer may mask large variations within individual cancer types -; not recognizing it may not help certain subgroups of patients.

New research from the University of Michigan Rogel Cancer Center aims to improve predictions of response to cancer drugs by distinguishing and allowing simultaneous examination of differences between several types of cancer as well as within individual types. The results appear in Computational biology PLOS.

It’s like the old argument between nature and education. Obviously, both are contributing. The questions we tried to answer were: What is each person’s contribution? And can we use this information to make predictions that would be useful in the clinic?“

Jun Li, Ph.D, co-lead author of the study and Professor of Human Genetics, Associate Research Chair in Computational Medicine and Bioinformatics, Michigan Medicine-University of Michigan

The study, led by former UM postdoctoral fellow John Lloyd, Ph.D., who had worked closely with Li and co-lead author Sofia Merajver, MD, Ph.D., used the MEK inhibitor response as an example of a proof of concept, and relied on two public datasets, each containing several hundred cancer cell lines derived from patients.

Analysis -; which included mRNA expression, point mutations and copy number variations -; found that while the predictions of drug response were very accurate when comparing one type of cancer as a whole to another type of cancer, the predictions only held for about five of the 10 types of cancer when looking at this type of cancer alone.

“This means that in order to be useful in the clinic for helping individual patients, we need to be able to incorporate inter-cancer and intra-cancer data,” adds lead co-author Matthew Soellner, Ph.D. assistant professor of chemistry at the University of Letters, Sciences and Arts of UM. “Otherwise, you can capture the average response to a drug for all types of cancer, but completely lose where an individual patient is in their type of cancer.”

For example, most colorectal cancer cell lines are susceptible to inhibition of MEK, but liver cancer cell lines show a much more mixed response. Thus, additional information and biomarkers beyond the single type of cancer would be important in determining the likelihood that a patient with liver cancer will respond well to an MEK inhibitor.

To this end, the UM team has developed a visualization strategy called a “cigar plot”, where the differences in response between cancer types can be visualized simultaneously with the responses within each cancer type. The longer the distribution of results, or the more cigar-shaped, within an individual cancer type, the better it can be used to predict the response in different people affected by that cancer type -; whether it is brain cancer or lung cancer.

“We hope this approach can serve as a general tool for the field,” says Merajver, professor of epidemiology and internal medicine. “More patients than ever are participating in basket clinical trials, which select based on the molecular characteristics of a cancer rather than where the cancer originated in the body, so prediction models will increasingly be more important in matching patients with effective treatments. “

This simultaneous approach to balancing considerations of both cancer type and individual variation within a type when making treatment decisions could also be applied to other diseases that affect multiple tissues or where predictions are drawn from large and diverse populations, the researchers note.