ABCP_finder: A Transformer Embedding-Based Prediction of Anti-Breast Cancer Peptides

Breast cancer is a leading cause of cancer-related deaths in women. Developing effective treatments for breast cancer faces challenges such as drug resistance, toxicity, and limited target specificity. Anti-breast cancer peptides (ABCPs) have shown promise as effective therapeutic candidates due to their low toxicity, high selectivity, and ability to target cancer cells.

Identifying novel ABCPs through experimental methods is a time-consuming and costly process. To tackle this issue, researchers developed ABCP_finder, a computational framework specifically designed for predicting ABCPs using transformer-based protein language model embeddings. The framework was built on positive and negative datasets that were carefully curated to ensure a biologically meaningful classification task.

To prevent data leakage and ensure realistic evaluation, a homology-aware train-test split strategy was employed using CD-HIT at 30% sequence identity with 80% coverage. Peptide representations were generated using pre-trained transformer models, ProtBERT and ESM2, and then classified using a multilayer perceptron (MLP). Among the tested models, ProtBERT showed superior performance, with an accuracy of 93.82%, a recall of 86.88%, an F1-score of 90.59%, a Matthews correlation coefficient (MCC) of 0.8618, an area under the receiver operating characteristic curve (AUC) of 96.67%, and a Brier Score of 0.0633.

These results demonstrate that ProtBERT has strong predictive capabilities, even in imbalanced conditions. A calibration analysis also supported the selection of a 0.7 probability threshold for identifying high-confidence ABCPs. Further external validation using xDeep-AcPEP showed that unknown peptide sequences predicted as ABCPs by ABCP_finder exhibit favorable inhibitory constant (IC) values, supporting their biological relevance.

Overall, ABCP_finder provides a reliable and practical platform for large-scale ABCP screening, potentially accelerating the discovery of novel peptide therapeutics for breast cancer treatment.