Integration of Protein Language Model Embeddings and Quantitative Structure-Activity Relationship Descriptors in a Gradient Boosting Framework for Antimicrobial Peptide Classification
DOI:
https://doi.org/10.31674/ijbb.2025.v03i02.001Abstract
In this work, we developed a hybrid eXtreme Gradient Boosting (XGBoost) framework for antimicrobial peptide (AMP) classification using a dataset of 20,968 sequences (2,001 AMPs). The framework combines embeddings from the ESM-2 protein language model with five quantitative structure-activity relationship (QSAR) features (net charge, hydrophobicity, length, positive residue count and hydrophobic residue ratio). Using five‑fold nested cross‑validation to avoid data leakage and scale_pos_weight to handle class imbalance, we conducted a systematic comparison of our hybrid model against three baselines: amino acid composition (AAC)-only, QSAR-only, and ESM-only. The hybrid model achieved an average AUC of 0.9137 and MCC of 0.6446, outperforming all three baselines. SHapley Additive exPlanations (SHAP) analysis identified sequence length as the most important feature, followed by several ESM-2 dimensions, with QSAR features playing a supporting role. This lightweight, interpretable framework offers a practical tool for computational AMP screening.
Keywords:
Antimicrobial Peptides, ESM‑2, QSAR Descriptors, XGBoost, Nested Cross‑validation, SHAP AnalysisReferences
Abbas, Munawar, et al. (2025) ABP-Xplorer: A Machine Learning Approach for Prediction of Antibacterial Peptides Targeting Mycobacterium abscessus-tRNA-Methyltransferase (TrmD). Journal of Chemical Information and Modeling, 65.11: 5456-5468.
Ali, Farman, et al. (2026) A generative explainable model for antimicrobial peptide prediction using bidirectional temporal convolutional neural network. Scientific Reports.
Bae, Daehun, et al. (2025) AI-guided discovery and optimization of antimicrobial peptides through species-aware language model. Briefings in Bioinformatics, 26.4: bbaf343.
Bale, Ashwin; Dutta, Arnab; Mitra, Debirupa. (2023) Combined charge and hydrophobicity-guided screening of antibacterial peptides: two-level approach to predict antibacterial activity and efficacy. Amino Acids, 55.7: 853-867.
Bhatnagar, Pranshul, et al. (2024) Predicting antibacterial activity, efficacy, and hemotoxicity of peptides using an explainable machine learning framework. Process Biochemistry, 145: 163-174.
Bin, Yannan, et al. (2025) Pepxml: ESM2-based extreme multilabel classification of pathogen-targeted antimicrobial peptides. Briefings in Bioinformatics, 26.5: bbaf548.
Bournez, Colin, et al. (2023) CalcAMP: A new machine learning model for the accurate prediction of antimicrobial activity of peptides. Antibiotics, 12.4: 725.
Brizuela, Carlos A., et al. (2025) AI methods for antimicrobial peptides: progress and challenges. Microbial Biotechnology, 18.1: e70072.
Chen, QingWei, et al. (2026) iAMP-SeE: an antimicrobial peptide recognition model based on ESM2 feature extraction and hybrid attention mechanisms. PeerJ, 14: e20978.
Cordoves-Delgado, Greneter; García-Jacas, César R. (2024) Predicting antimicrobial peptides using ESMFold-predicted structures and ESM-2-based amino acid features with graph deep learning. Journal of Chemical Information and Modeling, 64.10: 4310-4321.
García-González, Luis A., et al. (2025) Optimal Descriptor Subset Search via Chemical Information and Target Activity-Guided Algorithm for Antimicrobial Peptide Prediction. Journal of Chemical Information and Modeling, 65.13: 6621-6631.
Georgoulis, Elias; Zervou, Michaela Areti; Pantazis, Yannis. (2025) Transfer learning on protein language models improves antimicrobial peptide classification. Scientific Reports, 15.1: 37456.
Jullapech, Nawisa. (2025) Development and application of Machine Learning classification methods: optimal feature identification and prediction of antimicrobial peptides and other soft matter systems. 2025. PhD Thesis. University of Reading.
Lertampaiporn, Supatcha, et al. (2022) Ensemble-AHTPpred: a robust ensemble machine learning model integrated with a new composite feature for identifying antihypertensive peptides. Frontiers in Genetics, 2022, 13: 883766.
Lin, Changhang, et al. (2026) PepGraphormer: an ESM-GAT hybrid deep learning framework for antimicrobial peptide prediction. Journal of Cheminformatics, 18.1: 15.
Malshikare, Hrushikesh, et al. (2026) Mechanistic Principles of Antimicrobial Peptides Uncovered by Charge Density Based Machine Learning. Chemical Communications, 2026.
Mera-Banguero, Carlos, et al. (2024) AmpClass: an antimicrobial peptide predictor based on supervised machine learning. Anais da Academia Brasileira de Ciências, 2024, 96: e20230756.
Mohkam, Milad; Nezafat, Navid; Ghasemi, Younes. (2024) An in silico approach to de novo design of anti-microbial peptide from inspirited Komodo dragon's original VK6 peptide. International Journal of Data Mining and Bioinformatics, 2024, 28.1: 18-39.
Negovetić, Mario, et al. (2024) Efficiently solving the curse of feature-space dimensionality for improved peptide classification. Digital Discovery, 2024, 3.6: 1182-1193.
Pikalyova, Karina, et al. (2025) Design of Highly Potent Antibiofilm, Antimicrobial Peptides Using Explainable Artificial Intelligence. Journal of Chemical Information and Modeling, 2025, 66.1: 744-755.
Salimi, Abbas; Lee, Jin Yong. (2026) CoLPAT-AMP: A Transformer-Based framework for Designing novel antimicrobial peptides with property Awareness and partially controllable length. Expert Systems with Applications, 2026, 131999.
Shaon, Md Shazzad Hossain, et al. (2024) AMP-RNNpro: a two-stage approach for identification of antimicrobials using probabilistic features. Scientific Reports, 2024, 14.1: 12892.
Shoombuatong, Watshara, et al. (2025) Advancing the accuracy of anti-MRSA peptide prediction through integrating multi-source protein language models. Interdisciplinary Sciences: Computational Life Sciences, 2025, 17.3: 716-729.
Singh, Onkar. (2022) Integration of Machine Learning Along With Global Features of the Amino Acid Sequence in Predicting Antimicrobial and Anti-Inflammatory Peptides. 2022. PhD Thesis. National Yang Ming Chiao Tung University.
Teimouri, Hamid; Medvedeva, Angela; Kolomeisky, Anatoly B. (2023) Bacteria-specific feature selection for enhanced antimicrobial peptide activity predictions using machine-learning methods. Journal of Chemical Information and Modeling, 2023, 63.6: 1723-1733.
Yu, Qinze, et al. (2026) Uncovering evolutionarily remote and highly potent antimicrobial peptides with protein language models. Nature Biomedical Engineering, 2026, 1-14.
Zahid, Hamza, et al. (2026) PeptideNet: An Integrative Deep Learning Framework for Predicting Diverse Bioactive Peptides Using Protein Language Model Embeddings. Journal of Chemical Information and Modeling, 2026.
