Enhancing Anticancer Peptide Discovery: A Fusion-centric Framework with Conditional Diffusion for Prediction and Generation
Binyu Li,Xin Zhang, Zhihua Huang, Prayag Tiwari, Quan Zou, Yijie Ding, Xiaoyi Guo
Abstract:
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.
Introduction:
Cancer remains one of the leading threats to global public health, causing millions of deaths each year. Anticancer peptides (ACPs) are short peptides capable of selectively recognizing and killing cancer cells. Due to their high target specificity and low toxicity, ACPs have emerged as promising therapeutic candidates. With the rapid progress of peptide-based drug research, ACPs are increasingly regarded as effective complements to conventional therapies. A key advantage of ACPs lies in their ability to selectively target malignant cells while sparing healthy tissues.
Materials and Methods:
This study proposes UACD-ACPs, a unified framework for ACP classification and cancer-type-aware peptide generation. For classification, ProtBERT embeddings and physicochemical features are fused and refined using the multiscale enhanced convolutional structure (MECS) and a diffusion-inspired noise-conditioned encoder. The classification module and generation module share a common feature space, meaning both modules process the same set of features extracted from the dataset. While they share this feature space, each module operates independently during training, learning different representations based on their respective tasks.
Discussion:
This study presents a fusion-driven unified framework for anticancer peptide (ACP) discovery that integrates diffusion-inspired classification and diffusion-based sequence generation within a shared feature space. The architecture addresses key challenges in ACP modeling, including class imbalance, sequence heterogeneity, and the integration of predictive and generative objectives within a biologically coherent framework.
Citation: Li B, Zhang X, Huang Z, Tiwari P, Zou Q, Ding Y, et al. (2026) Enhancing anticancer peptide discovery: A fusion-centric framework with conditional diffusion for prediction and generation. PLoS Comput Biol 22(3): e1014098. https://doi.org/10.1371/journal.pcbi.1014098
Editor: Jeffrey Skolnick, Georgia Institute of Technology, UNITED STATES OF AMERICA
Received: November 19, 2025; Accepted: March 5, 2026; Published: March 26, 2026.
Copyright: © 2026 Li et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data and code can be accessed from https://github.com/yidingneng/ACP-ConditionalDiffusion.git.
Funding: This work was supported by the National Natural Science Foundation of China (No. 62450002 to Q.Z.; No. 62302340 to X.G.; U22A2038 to Y.D.), and the Municipal Government of Quzhou (Grant No. 2024D002 to Y.D.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
