Reply to: The pitfalls of negative data bias for the T-cell epitope specificity challenge

Abstract

Predicting and identifying TCR-antigen pairings accurately presents a significant computational challenge within the field of immunology. The negative sampling issue is important T-cell specificity modeling and it is known clearly by the community that different negative data sampling strategy will influence the prediction results. Therefore, proper negative data sampling strategy should be carefully selected, and this is exactly what PanPep has noticed, emphasized and performed. Now we would like to clarify this point further by formulating this problem as a PU learning. Our findings suggest that the reshuffling strategy may generate potential false negative samples, which can adversely affect model training and result in biased model testing for PanPep. Furthermore, a proper comparison between different negative sampling strategies should be performed in a consistent way to make a proper conclusion. Finally, future updating to explore more possible and suitable negative sampling strategy is expected.