LysePred: A Multiscale Convolutional Neural Network for Predicting Hemolytic Activity of Antimicrobial Peptides

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Mean MCC improvement of 9.13% and mean ACC improvement of 3.65% over the next-best comparator; single-scale ablation variants showed up to 13.35% MCC degradation, quantifying the multiscale architecture contribution

• Primary Outcome: LysePred exceeded the second-best method by 9.13% in MCC and 3.65% in ACC on average across six benchmark datasets; independent validation on HemoPI2 demonstrated highly competitive performance

• Nnt Or Sensitivity: Sensitivity and specificity not explicitly reported in abstract; MCC CV < 6.92% and ACC CV < 2.73% across cross-validation folds indicate model stability; parameter count ~0.55M vs. 8M (ESM-2) and 110M (BERT Base) for computational efficiency comparison

• Confidence Interval: Not reported

Clinical Application

High feasibility as a research tool: the model is computationally lightweight (~0.55M parameters, linear complexity), publicly available via GitHub, and requires only amino acid sequence input. Integration into AMP design pipelines is straightforward. However, wet-laboratory confirmation of predictions remains mandatory before any candidate advances to preclinical development. Australia faces significant antimicrobial resistance (AMR) burden, with the ACSQHC and the National Antimicrobial Resistance Strategy 2020–2025 prioritising novel antibiotic development. LysePred could support Australian academic and pharmaceutical research groups engaged in AMP discovery by enabling rapid computational triage of hemolytic candidates prior to costly in vitro assays. However, no AMP has yet achieved TGA registration, and PBS listing is not applicable at this stage. RACGP guidelines do not currently address AMP therapeutics. Any Australian research application would require institutional biosafety and ethics oversight for subsequent experimental validation phases. The tool does not replace TGA-required preclinical toxicology packages. LysePred is applicable as an early-stage in silico screening tool for researchers designing or selecting AMPs for antibiotic development programmes. It is not directly applicable to patient care at this stage. Potential end-users are pharmaceutical scientists, synthetic biologists, and medicinal chemists engaged in AMP lead optimisation.

Abstract

Antimicrobial peptides (AMPs) represent promising alternatives to conventional antibiotics, yet hemolytic toxicity remains a critical barrier to clinical translation, with approximately 70% of known AMPs exhibiting high or moderate hemolytic activity. Existing computational prediction methods are often constrained by deficiencies, including high computational complexities, the inability to capture multiscale sequence patterns, and insufficient generalization across diverse datasets. We present LysePred, a multiscale convolutional neural network to address these deficiencies concurrently by employing parallel branches with exponentially spaced kernel sizes to simultaneously capture local amino acid motifs (bigrams, 4-g) and longer-range amphipathic patterns (8- to 32-g). LysePred achieves top-tier performance on six benchmark datasets, exceeding the performance of the second-best method by 9.13% in MCC and 3.65% in ACC on average, while maintaining exceptional stability (MCC CV < 6.92%, ACC CV < 2.73%). Furthermore, independent validation on the HemoPI2 dataset demonstrates that LysePred delivers highly competitive results with a computationally parsimonious design (∼0.55 M parameters). This represents a reduction in parameter density of 1 to 2 orders of magnitude compared to Transformer-based approaches, such as the 8M-parameter ESM-2 or the 110M-parameter BERT Base while maintaining linear complexity for high-throughput screening. Ablation studies validate that the multiscale architecture contributes meaningfully to performance, with single-scale variants showing up to 13.35% MCC degradation. Interpretability analyses via t-SNE visualization and SHAP feature importance reveal that LysePred learns biologically meaningful representations integrating both local sequence motifs and global structural patterns. LysePred offers a practical, efficient, and interpretable tool for rapid hemolytic toxicity prediction in antimicrobial peptide development. Code and data are available at https://github.com/lincubator/LysePred.

References

1. 1.Lin, C., Li, J., Su, C., Xiong, S., Kang, X., Lu, J., & Wei, L. (2026). LysePred: A multiscale convolutional neural network for predicting hemolytic activity of antimicrobial peptides. ACS Synthetic Biology. Advance online publication. https://doi.org/10.1021/acssynbio.6c00173