At London's Institute of Cancer Research, scientists have quietly developed a tool that could spare thousands of advanced bowel cancer patients from drugs that won't help them—and may harm them. The breakthrough is called PhenMap, an AI system that uses genetic data to predict which patients will actually respond to bevacizumab, a newly approved NHS drug, before they receive it.
The stakes are high. Nearly 10,000 cases of advanced bowel cancer are diagnosed in the UK each year, and the disease carries an unforgiving prognosis. While survival rates for early-stage bowel cancer can reach 98%, the five-year survival rate for advanced cases drops to as low as 10 percent. Bowel cancer has the second highest mortality rate of any cancer, behind only lung cancer. When patients reach this stage, treatment options are desperately limited.
Bevacizumab, approved by the NHS in December, works by starving tumours of the proteins they need to grow. It sounds promising on paper—but here's the problem: it only helps a small fraction of patients, and carries serious side effects including blood clots and gastrointestinal complications. This means that without a way to identify who will benefit, thousands of patients could endure those unpleasant effects for no therapeutic gain. Researchers at the Institute of Cancer Research and the RCSI University of Medicine and Health Sciences in Dublin set out to change that.
The team studied 117 European bowel cancer patients who had been treated with chemotherapy and bevacizumab. Using PhenMap—a portmanteau of "phenotype" (an organism's observable traits) and "mapping"—they integrated complex genetic data about each patient's tumour. The AI tool spotted patterns in how different patients responded to the drug, patterns that would be invisible to human analysis. Most strikingly, it identified a group of patients who all shared the same gene mutation and faced a high risk of negative reactions.
Professor Anguraj Sadanandam from the Institute of Cancer Research framed the finding in terms that cut to the heart of the matter: "We know that the majority of patients won't benefit from the drug, meaning thousands of people in England could be facing unpleasant side effects unnecessarily. Until now, we haven't been able to identify these patients." Now they can—or at least, the foundation exists to do so.
The researchers acknowledge that PhenMap needs validation on a larger patient sample before it can be deployed in clinics. But the vision is clear. Sadanandam said he hopes the approach will eventually lead to a test that clinicians can use routinely, ensuring patients receive "personalised care that has the highest chance of working against their cancer." The team also plans to explore whether the method works for other cancer types.
What makes this particularly significant is that it exemplifies a broader shift in medicine: using AI not to replace human judgment, but to reveal patterns in data so complex that only machines can see them. The promise isn't just better outcomes—it's the chance to spare patients needless suffering by targeting treatment to those most likely to benefit. For thousands of people facing advanced bowel cancer each year, that distinction could matter enormously.