Sophisticated new computer software can be used to predict how cancers may respond to a new drug before it has ever been given to patients.

The new program could transform the discovery of cancer drugs by predicting how tumors will become resistant to treatment long before it would first become apparent in clinical trials.

Based on the software’s predictions, researchers could start working on second-generation drugs to tackle treatment resistance before the first-generation drug is taken to patients.

It could also lead to the development of tests to assess patients for resistance mutations before and during treatment—delivering precision medicine at the earliest stage.

The new prediction tool was developed by scientists at The Institute of Cancer Research, London, with funding from Cancer Research UK, and is described in the journal Cell Chemical Biology today.

First prediction tool to include evolutionary impact

The approach begins by analyzing all the possible mutations that could occur around a drug target—generally between 350 and 1,200.

The researchers then apply the prediction software to prioritize the mutations down to only 9 or 10 most likely to cause drug resistance—a more feasible number to investigate further in the laboratory.

The researchers tested their method on existing cancer drugs and drug targets—including 17 different drugs that target the important cancer-related proteins MAPK1, KIT, EGFR, Abl, and ALK. It was able to accurately predict many of the mutations that doctors see in the clinic, and for MAPK, many that were generated in the lab.

The prediction tool is the first to include the evolutionary impact of a mutation on cancer cells. If a mutation meant the drug target could no longer perform its role in a cell, then that cell is unlikely to survive and go on to form drug-resistant tumors.

Mutations could be either single- or double-letter changes in genes that would lead to a change in the building blocks (known as amino acids) that make up a protein.

Mutations also had to be in close vicinity to the site where the drug binds its target, and had to affect the drug target in a way that means the drug binds less tightly.

Lastly, the prediction tool identifies regions in the drug target where resistance ‘hotspots’—areas predicted to have multiple mutations—are likely to occur and prioritized mutations at these ‘hotspots’ based on their likelihood of being formed in the cancer type under investigation.