Year: 2026
Type: preprint
Abstract: Breast cancer remains one of the leading causes of cancer-related deaths among women worldwide. Anti-breast cancer peptides (ABCPs) have emerged as effective drug candidate due to its low toxicity, high selectivity, and ability to target cancer. However, it is very time consuming and expensive to identify novel ABCPs only through experiment methods. To address this challenge, we developed ABCP_finder, the first dedicated computational framework specifically designed for the prediction of ABCPs. The curated dataset comprised of 301 experimentally validated ABCPs and 1052 non-ABCPs. Unlike general anticancer peptide (ACP) predictors which are not trained on breast cancer specific datasets. It was trained using deliberately stringent and biologically relevant ABCPs and non-ABCPs. Instead of random train-test split, homology-aware split using CD-HIT was performed to prevent data leakage. Peptide representations generated using ProtBERT and ESM2 were classified using a multilayer perceptron (MLP). Among the tested models, ProtBERT-based model showed superior performance, achieving 93.82% accuracy, 86.88% recall, 90.59% F1-score, 0.8618 MCC, 96.67% AUC, and a Brier Score of 0.0633. Calibration analysis supported the selection of a 0.7 probability threshold for identifying high confidence ABCPs. Further external validation using xDeep-AcPEP demonstrated that peptide predicted as ABCPs by ABCP_finder are exhibiting favourable IC□□ values. ABCP_finder provides a reliable platform for high throughput large-scale ABCP screening of novel peptide therapeutics for breast cancer treatment. Abstract Figure Caption: Figure illustrating the graphical abstract of ABCP_finder.
Language: English
Subfield: Drug Discovery
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7212835078