Breakthrough in Precision Oncology: ICR's AI Tool Transforms Bowel Cancer Treatment
Bowel cancer, medically termed colorectal cancer, remains one of the most prevalent malignancies in the United Kingdom, claiming thousands of lives annually. With approximately 46,600 new diagnoses each year and around 17,000 deaths, it ranks as the fourth most common cancer and the second leading cause of cancer mortality. Survival rates drop dramatically for advanced stages, where the cancer has metastasized, hovering at just 10% five-year survival. This stark reality underscores the urgent need for tailored therapies that maximize efficacy while minimizing harm.
Recently, scientists at The Institute of Cancer Research (ICR), London—a premier postgraduate research university and member of the University of London—have pioneered an artificial intelligence (AI)-driven method to predict patient responses to bevacizumab, a key targeted therapy now available on the National Health Service (NHS). This innovation promises to usher in a new era of personalized bowel cancer treatment, ensuring patients receive interventions suited to their unique tumor biology.
The Challenge of Advanced Bowel Cancer and Standard Treatments
Advanced bowel cancer, or metastatic colorectal cancer (mCRC), occurs when tumors spread beyond the colon or rectum to distant organs like the liver or lungs. Nearly 10,000 such cases are diagnosed annually in England alone, with incidence rising alarmingly among younger adults under 50. Traditional chemotherapy offers limited success, prompting the integration of targeted drugs.
Bevacizumab (Avastin), a monoclonal antibody inhibiting vascular endothelial growth factor (VEGF), starves tumors of blood supply when combined with chemotherapy. Approved by the National Institute for Health and Care Excellence (NICE) in December 2025 under guidance TA1136 for first- and second-line mCRC treatment, it extends progression-free survival for some. However, only 10-20% of patients achieve meaningful benefit, while others endure severe side effects—high blood pressure, gastrointestinal perforations, and thrombosis—without gain.
- Effective in restricting angiogenesis but fails in resistant cases due to tumor heterogeneity.
- Thousands potentially exposed unnecessarily, straining NHS resources.
- Need for biomarkers to stratify responders pre-treatment.
Introducing PhenMap: ICR's Revolutionary AI Predictor
The PhenMap tool, developed by ICR's Professor Anguraj Sadanandam and collaborators at RCSI University of Medicine and Health Sciences in Dublin, fuses multi-modal data—genomic copy number aberrations (CNAs), mutations, and clinical variables like age, gender, and tumor location—into actionable insights. Detailed in a April 2026 Scientific Reports paper (DOI: 10.1038/s41598-026-39189-w), it employs sparse Bayesian factor analysis to identify 'mapping variables' (MVs) capturing hidden tumor phenotypes.
The process unfolds step-by-step:
- Data Integration: Compile CNAs (via GISTIC scores), somatic mutations (e.g., BRAF), and clinical covariates from patient cohorts.
- PhenMap Analysis: Bayesian modeling extracts 10 key MVs; two (MV5, MV8) link to progression-free survival (PFS), tied to 15q21.1/1p36.31 deletions and BRAF mutations.
- Risk Stratification: Elastic net Cox regression yields a prognostic score, grouping patients: low-risk (bottom 10%, 88% responders), moderate (middle 80%), high-risk (top 10%, 0% responders).
- Validation: Cox models confirm independent PFS prediction (C-index 0.60); AUC 0.685 for response.
Validated on 117 bevacizumab-treated mCRC patients from the ANGIOPREDICT cohort, high-risk cases showed 100% non-response and 10.78-fold higher mortality hazard versus low-risk.
Behind the Innovation: ICR's Research Ecosystem
The ICR, ranked top in UK for biological sciences research impact, exemplifies higher education's role in translational oncology. As a specialist university, it trains PhD students and postdocs in AI, genomics, and precision medicine, fostering collaborations like this EU Horizon 2020-funded effort. Professor Sadanandam, in ICR's Division of Cancer Biology, leads stratification efforts, emphasizing multi-omics fusion for 'hidden clues' in tumors.
Funding from Research Ireland and ICR trusts enabled this, highlighting public-charity partnerships driving UK research excellence.
Photo by National Cancer Institute on Unsplash
Patient Impact: Sparing Unnecessary Suffering and Costs
For the ~10,000 annual advanced cases in England, PhenMap could identify non-responders early, avoiding bevacizumab's toxicities and enabling alternative therapies. Professor Sadanandam notes: 'Thousands could face unpleasant side effects unnecessarily... Our AI spots patterns impossible for humans.'
NHS savings from targeted use align with NICE's value-based approvals. Patients gain confidence in personalized plans, improving quality of life amid rising young-adult diagnoses.
Broader Implications for Precision Medicine in the UK
This advances UK's 100,000 Genomes Project legacy, integrating AI into oncology. BRAF mutations, flagged as high-risk biomarkers, open doors to combination trials (e.g., EGFR inhibitors). Scalable to other cancers, PhenMap embodies ICR's vision for 'smarter, kinder therapies,' per CEO Professor Kristian Helin.
ICR's announcement details expansion plans.
Future Directions: From Bench to Bedside
Next: larger validation cohorts, prospective trials, clinical assay development. Potential for liquid biopsies or histopathology integration. ICR eyes multi-cancer applications, aligning with NHS AI Lab initiatives.
- Prospective NHS trials for bevacizumab selection.
- New drugs targeting high-risk pathways (e.g., BRAF/MAPK).
- AI training datasets from UK biobanks.
Training the Next Generation: Careers in Cancer AI Research
ICR's postgraduate programs equip researchers for precision medicine. PhDs in bioinformatics, oncology fuse AI with biology, vital amid UK's 7% projected incidence decline yet persistent advanced burdens. Opportunities abound in computational oncology, with demand for data scientists in research jobs.
Photo by National Cancer Institute on Unsplash
Stakeholder Perspectives: Patients, Clinicians, Policymakers
Patient advocates hail personalization; clinicians anticipate reduced trial-and-error. Policymakers eye cost efficiencies amid NHS pressures. Balanced views note need for diverse data to avoid biases.
Real-world case: BRAF-mutant patients rerouted to immunotherapy trials, extending survival.
Global Context and UK Leadership
While Singapore's CAN-Scan explores similar ML for CRC, ICR's PhenMap leads in bevacizumab stratification. UK's genomics prowess positions it forefront precision oncology.
Actionable insights: Advocate screening, support research funding.