Extended Data Fig. 1: Predictive performance of PISTE using unified-peptide negative sampling and reference-TCR negative sampling methods. | Nature Machine Intelligence

Extended Data Fig. 1: Predictive performance of PISTE using unified-peptide negative sampling and reference-TCR negative sampling methods.

From: Sliding-attention transformer neural architecture for predicting T cell receptor–antigen–human leucocyte antigen binding

Extended Data Fig. 1

All testing triples whose Antigen-HLA pairs were observed in the training data are removed from the test-sets. (a) The AUROC, AUPR and PPVn for PISTE and competing models using the unified-peptide negative sampling schemes. (b) The AUROC, AUPR and PPVn for PISTE and competing models using the reference-TCR negative sampling schemes. The red baseline represents a random classifier.

Back to article page