ImmuneWatch DETECT Model Leads the Field in IMMREP23 TCR-epitope Prediction Challenge

ImmuneWatch is proud to announce that its proprietary TCR-epitope prediction tool, ImmuneWatch DETECT, has achieved top performance in the IMMREP23 benchmark competition. IMMREP23, the second major benchmark of its kind, brought together global experts in immunology and machine learning to assess the accuracy of various models in predicting the specificity of T-cell receptors (TCR) towards peptide-MHC (pMHC) targets.

The challenge addressed a critical question for the future of immune monitoring: can we reliably predict the target of a T-cell receptor? Over 50 teams submitted nearly 400 solutions to tackle this problem, testing their methods against a carefully curated dataset. The results from the benchmark underscored several important lessons:

• Predictive Accuracy with Sufficient Data: Models performed well when they had access to extensive training data on specific (“seen”) epitopes.
• Challenges with Novel Epitope Prediction: Predicting interactions for new (“unseen”) epitopes remains an unsolved issue in the field.
• Importance of Competition Design: The presence of target leakage in the test set, introduced by the method chosen by the organisers for simulating negative data, complicated the ranking of methods. This is a valuable lesson learned for a future edition.

Among the methods tested, ImmuneWatch’s DETECT stood out with what the final manuscript described as a “substantial predictive advantage,” outperforming other models in its category. This achievement reflects years of development and dedication from the ImmuneWatch team to build a robust tool capable of supporting innovative T-cell research.

ImmuneWatch CEO, Sander Wuyts, commented on the company’s participation in IMMREP23, stating:

“We approached IMMREP23 with a mixture of excitement and apprehension. While there was a risk, it was also a valuable opportunity to learn and grow. Seeing ImmuneWatch DETECT recognised as a top performer validates our work and motivates us to tackle the ongoing challenges, especially around ‘unseen’ epitopes.”

Looking forward, ImmuneWatch is committed to refining ImmuneWatch DETECT and expanding its capabilities. The goal is to make groundbreaking strides in TCR specificity prediction, providing the field with tools to better understand immune responses and accelerate the development of personalised therapies.

The full manuscript detailing the findings from IMMREP23 is available here: Read the full paper

About ImmuneWatch

ImmuneWatch specialises in immune system monitoring through advanced machine learning tools. By predicting TCR-epitope interactions, ImmuneWatch supports researchers, biotech companies, and pharmaceutical innovators in developing more effective immunotherapies and vaccines.

‍