CD47BPFormer: A Sequence-Based Predictor Based on Transformer to Identify CD47 Binding Peptides

The CD47/SIRPα pathway, like the PD1/PDL1 pathway, has been shown to play a crucial role in cancer immunotherapy. However, monoclonal antibody therapies targeting this pathway have inherent limitations. To overcome these limitations, peptide drugs can be used as a substitute for monoclonal antibody therapies. In this study, we used the CD47 binding peptide isolated by next-generation phage display as a benchmark dataset to train a predictor based on transformer called CD47BPFormer. Our results show that CD47BPFormer can effectively identify CD47 binding peptides. In the five-fold cross-validation, CD47BPFormer achieved an accuracy of 0.800, a sensitivity of 0.750, a specificity of 0.850, a Matthew's correlation coefficient of 0.603, an F-score of 0.790, a precision of 0.833, an area under the receiver operating characteristic curve of 0.825, and an area under the precision-recall curve of 0.833.