06.05.2026

Introducing BioCOMPASS at ICLR 2026

In April 2026, I presented a paper in a workshop called Machine Learning for Genomics Exploration (MLGenX) at the The Fourteenth International Conference on Learning Representations (ICLR) 2026.

The workshop focused on bridging machine learning and genomics and themed around AI systems that can reason, experiment, and learn from biological feedback this year. My poster introduced BioCOMPASS, an extension of the COMPASS transformer model for predicting how patients respond to cancer immunotherapy. The core idea is to ground biological interpretability by integrating external clinical biomarkers and treatment information into the model’s learning process. This also helped the model generalise better across different cancer types, drugs, and patient cohorts. It was a valuable experience to engage with others working at the same intersection.

oplus_32