When it comes to cancer treatment, timing and precision can mean everything. But even today, many patients are matched with therapies through trial and error, enduring months of ineffective treatments. What if doctors could predict from the outset how some people might respond, based on the unique signature of their genes?

A groundbreaking study led by Ruishan Liu, USC Viterbi assistant professor of computer science, may bring that vision closer to reality. Liu and a team of researchers, in collaboration with Genentech and Stanford University, analyzed genomic and clinical data from more than 78,000 patients across 20 cancer types — making it the largest study of its kind to date.

Their findings, published in the journal Nature Communications, identified nearly 800 genetic alterations that influence treatment outcomes, including 95 genes significantly associated with survival in breast, ovarian, skin and gastrointestinal cancers.

“These discoveries highlight how genetic profiling can play a crucial role in personalizing cancer care,” Liu said. “By understanding how different mutations influence treatment response, doctors can select the most effective therapies — potentially avoiding ineffective therapies and focusing on those most likely to help.”

Among the study’s most compelling revelations: Certain mutations have a dramatic impact on how patients respond to different types of treatment, such as chemotherapy, immunotherapy or targeted drugs.

For instance, KRAS (Kirsten rat sarcoma virus) mutations — common in lung and colorectal cancers — were linked to poor responses to EGFR (epidermal growth factor receptor) inhibitors. Meanwhile, mutations in the NF1 (neurofibromatosis type 1) gene improved responses to immunotherapy but reduced effectiveness of some targeted treatments. Others, like DNA repair pathway mutations, made tumors more unstable and therefore more vulnerable to immunotherapy.

In short, the presence, or absence, of a particular mutation could change everything.

To build on these insights, Liu’s team developed a machine learning tool to predict how patients with advanced lung cancer might respond to immunotherapy, integrating the effects of multiple mutations at once.

“Our goal was to find patterns that might not be obvious at first glance,” Liu said.

While precision medicine has long been heralded as the future of oncology, relatively few genetic mutations currently have treatments linked to them. Liu’s study helps fill in the gaps, offering clinicians a richer map to navigate a disease that is as varied as the patients who face it.

“By moving beyond a one-size-fits-all approach,” Liu said, “we can start to tailor therapies in ways that give patients the best chance at better outcomes.”

The study’s coauthors include Shemra Rizzo, Lisa Wang, Nayan Chaudhary, Sophia Maund, Sarah McGough, and Ryan Copping of Genentech; Marius Rene Garmhausen of Roche; and James Zou of Stanford University.