Enhancing anticancer peptide discovery: A fusion-centric framework with conditional diffusion for prediction and generation

Anticancer peptides (ACPs) are short bioactive sequences that selectively target tumor cells with minimal toxicity, positioning them as promising candidates for next-generation cancer therapies. However, existing computational models face limitations in sequence representation and class imbalance. To address these challenges, we propose UACD-ACPs, a unified fusion-driven framework that integrates a diffusion-inspired noise-conditioned classifier for ACP prediction and a diffusion-based peptide generation module with cancer-type-aware organization for targeted downstream screening. The classification module integrates ProtBERT-based semantic embeddings with physicochemical descriptors via the Multiscale Embedding Compression Strategy (MECS) and a diffusion-inspired noise-conditioned encoder, substantially enhancing predictive robustness and accuracy, particularly under challenging imbalanced multi-class settings. In the generative pipeline, we introduce a denoising diffusion-based generative framework augmented by two novel fusion modules: the Bitemporal Fusion Module (BFM) and the Temporal Feature Attention Module (TFAM). These modules perform multi-scale temporal and semantic fusion to promote the generation of structurally coherent and functionally relevant peptide candidates. Experimental results demonstrate that UACD-ACPs outperforms state-of-the-art methods in terms of accuracy, F1-score, and AUC-ROC. The generated peptides exhibit favorable physicochemical properties, diverse secondary structures, and strong structural stability, as validated by molecular dynamics simulations and membrane-binding analyses. Overall, this study highlights the potential of fusion-driven diffusion-based frameworks for alleviating class imbalance and data heterogeneity in anticancer peptide modeling, paving the way for scalable and biologically grounded ACP discovery.Author summary: Anticancer peptides (ACPs) are short protein sequences that can selectively target tumor cells while causing minimal harm to healthy tissues, making them promising candidates for cancer therapy. However, the computational discovery of ACPs remains challenging because peptide sequences are highly diverse and data across different cancer types are often severely imbalanced. In this study, we propose UACD-ACPs, a unified computational framework designed to support both the prediction and generation of anticancer peptides. By integrating complementary sequence representations with physicochemical information, the framework improves the reliability of ACP identification under challenging imbalanced data conditions. In addition, the model enables the generation of novel peptide candidates organized by cancer type, facilitating targeted downstream screening. Experimental results demonstrate that UACD-ACPs achieves improved predictive performance compared with existing methods. The generated peptides also exhibit favorable physicochemical properties and structural stability, suggesting their potential biological relevance. Overall, this work provides an integrated and scalable computational strategy to support anticancer peptide discovery and guide future experimental studies.