Researchers at IRB Barcelona, based at the Barcelona Science Park, have developed DiffInvex, a computational marker that allows researchers to study how the evolutionary pressure on genes changes as healthy cells transform into tumour cells, and how the latter evolve after exposure to chemotherapy. The study, published in the journal Nature Communications, has analysed more than 8,500 complete tumour genomes, identifying 11 genes associated with resistance to specific anticancer drugs.

Just as species adapt over generations, our body’s cells accumulate DNA changes throughout life. Most are harmless, yet a few “driver” mutations give a cell a competitive edge and can spark cancer. Chemotherapy then adds a new evolutionary pressure, encouraging further genetic changes that let tumours bounce back.

Chemotherapy remains a cornerstone of cancer therapy, yet many tumours eventually relapse. Deciphering which mutations help cells survive treatment is notoriously difficult because chemotherapy itself causes new DNA damage and patients often receive a cocktail of drugs. “We needed a way to see through that noise and catch evolution in the act,” says senior author Dr. Fran Supek, group leader at IRB Barcelona and professor at the Biotech Research & Innovation Centre (BRIC), University of Copenhagen.

A data-driven framework powered by whole-genome sequencing

DiffInvex empirically infers a “neutral” mutation rate baseline for important, coding gene regions, by contrasting them with mutations in adjacent non-coding regions such as introns or intergenic regions. This removes the guesswork from the assessment of how different factors influence mutation rates and spectra during tumour evolution and chemotherapy.

Drawing on more than 11,000 human genomes from roughly 30 tissue types, DiffInvex has identified 11 genes whose mutations are favoured more strongly after exposure to specific types of chemotherapy, implicating well-known drivers such as PIK3CA, SMAD4 and STK11. These findings suggest that resistance to anticancer drugs is often mediated by the accumulation of additional driver mutations in known cancer genes rather than by specialized mutations in specific drug-resistance genes.

The study also compared 1,722 genomes from healthy tissues to matched tumour types and shows that mutations in ARID1A—long considered a tumour-suppressor driver—and in other cancer genes are frequently selected during normal ageing. These observations would therefore suggest that some so-called cancer drivers may be evolutionary relics rather than disease initiators. “Our work reveals that cancer’s favourite strategy is not building bespoke shields against a particular drug, but rather boosting its core circuitry so that (almost) any blow hurts less,” says Dr. Supek.

Implications for precision oncology

The identification of “generalist” resistance paths opens the door to rational drug combinations: pairing standard chemotherapy with inhibitors that block PIK3CA or STK11 signalling, for example, might delay or prevent relapse. Meanwhile, recognition that some apparent driver mutations (such as ARID1A) pre-date cancer could improve early-detection panels and spare patients unnecessary worry.

“By disentangling treatment effects from background noise, DiffInvex could one day help clinicians predict the resistance pathways a patient’s tumour is likely to take – and cut them off in advance,” concludes Dr. Ahmed Khalil, first author of the study, a former postdoctoral fellow at IRB Barcelona, now a senior data scientist at IMIDomics, a biotech research company also located at the Barcelona Science Park.

» Article of reference: DiffInvex identifies evolutionary shifts in driver gene repertoires during tumorigenesis and chemotherapy. Ahmed Khalil, Fran Supek. Nature Communications (2025). doi: 10.1038/s41467-025-59397-8

» Link to the news: web de l’IRB Barcelona [+]