Anti-angiogenic agents are cornerstones in the treatment of patients with metastatic colorectal cancer. However, predictive biomarkers for the efficacy of these agents are still missing. In this project, we analyzed tumor and blood samples as well as computed tomography (CT) scans from patients receiving chemotherapy with or without an anti-angiogenic agent within the randomized RAMTAS trial. We analyzed gene expression profiles, post-translational protein modifications and somatic gene mutations in the primary tumor and in circulation free DNA (cfDNA). CT scans were analyzed using artificial intelligence (AI) with deep-learning radiomics approaches. Within this project primary tumor- and serum/ plasma samples from more than 350 patients and over 800 CT scans were collected and analyzed.

Interestingly, we observed different glycosylation of immunoglobulins in plasma samples and differently expressed N-glycans in primary tumor samples between patients receiving chemotherapy or a combination with the anti-angiogenic agent. In cfDNA analyses, we identified an over-expression in EGFR and MET gene copy number gain in about 50 % of patients. Those patients with higher EGFR copy number gains were less likely to benefit from treatment.

This results could be validated in an independent cohort treated with a different anti-angiogenic agent. Interestingly, EGFR copy number gains were higher in chemotherapy-refractory patients compared with patients treated in earlier lines. Finally, the complex interplay between these predictive and prognostic signatures as well as the radiologic parameters will be explored and validated using an already established AI based algorithm as soon as the final results of ongoing RAMTAS trial will be available.

In summary, this results will help to identify colorectal cancer patients with the most likelihood of response to anti-angiogenic therapies and identify potential resistance mechanism for further development of systemic treatment in these patients.