BioCOMPASS: Integrating Biomarkers into Transformer-Based Immunotherapy Response Prediction

TL;DR

BioCOMPASS enhances transformer-based immunotherapy response prediction by integrating biomarkers through novel loss components, improving generalisability across diverse datasets and treatment scenarios.

Contribution

It introduces a new method that incorporates biomarkers into transformer models via loss components, leading to better generalisation in immunotherapy response prediction.

Findings

Components like treatment gating and pathway consistency loss improve model generalisability.

BioCOMPASS outperforms previous models in leave-one-cohort/cancer/treatment-out evaluations.

Integrating biomarkers via loss functions enhances model robustness across diverse datasets.

Abstract

Datasets used in immunotherapy response prediction are typically small in size, as well as diverse in cancer type, drug administered, and sequencer used. Models often drop in performance when tested on patient cohorts that are not included in the training process. Recent work has shown that transformer-based models along with self-supervised learning show better generalisation performance than threshold-based biomarkers, but is still suboptimal. We present BioCOMPASS, an extension of a transformer-based model called COMPASS, that integrates biomarkers and treatment information to further improve its generalisability. Instead of feeding biomarker data as input, we built loss components to align them with the model's intermediate representations. We found that components such as treatment gating and pathway consistency loss improved generalisability when evaluated with Leave-one-cohort-out, Leave-one-cancer-type-out and Leave-one-treatment-out strategies. Results show that building components that exploit biomarker and treatment information can help in generalisability of immunotherapy response prediction. Careful curation of additional components that leverage complementary clinical information and domain knowledge represents a promising direction for future research.