An Interpretable Multi-Objective Machine Learning Framework for In Silico Prioritization of Anti-Staphylococcus aureus Antimicrobial Peptides

Machine Learning Model for Predicting Antibiotic Resistance Patterns from Protein Sequences

Antibiotic resistance (AR) has emerged as a pressing global health challenge, undermining the effectiveness of conventional treatment options and threatening public health systems around the world. The rapid identification of resistance genes and their associated mechanisms is therefore critical for the development of diagnostic and therapeutic strategies. This study presents the development of an ensemble machine learning pipeline for predicting resistance gene functions using the Comprehensive Antibiotic Resistance Database (CARD). Protein sequences were extracted from CARD and processed into features using sequence-derived representations, including kmer embeddings and TF-IDF vectorization. An ensemble voting classifier was implemented, combining five base estimators: Extra Trees Classifier, Random Forest, XGBoost, Linear Discriminant Analysis, and K-Nearest Neighbors. The ensemble approach utilized both hard and soft voting strategies, with optimized weights determined through log-loss minimization. The system was evaluated on a curated subset of the CARD dataset, ensuring balanced class representation across 62 antibiotic drug classes. Results demonstrate that the ensemble approach achieves superior classification performance, with the optimized soft voting classifier achieving 89.76% accuracy, 92% precision, 90% recall, and 90% F1-score. The hard voting ensemble achieved 89.44% accuracy with comparable precision and recall metrics. These results represent significant improvements over individual base classifiers, highlighting the effectiveness of ensemble methods for antibiotic resistance prediction. The ensemble approach demonstrates superior performance while maintaining computational efficiency, making it suitable for deployment in resourceconstrained environments.

Coarse composition suffices: tabular in-context learning for multi-activity antimicrobial peptide profiling

Antimicrobial peptides (AMPs) often act against multiple pathogen classes, making multi-label activity prediction a more realistic screening target than binary antimicrobial classification. The ESCAPE benchmark formalizes this setting, but leading approaches typically rely on multimodal, structure-conditioned deep models that are costly to train and tune. We show that a simple, sequence-only pipeline can match and surpass these methods by combining 330 interpretable sequence descriptors with TabPFN, a tabular foundation model that performs in-context prediction in a single forward pass without gradient-based training or hyperparameter search. On ESCAPE (82,359 peptides; five labels), a label-powerset TabPFN model achieves mAP-5 = 77.8%, improving on the previously best reported 72.1%. A probabilistic classifier chain is the first method to match or exceed the best published average precision on each of the five labels simultaneously. The gains persist under the prior state-of-the-art single-fold training protocol, indicating they are not a training-set-size artefact, and are largest for remote homologues (+11.2 points below 30% sequence identity). Ablations further show that predicted structure is unnecessary at inference and that performance is not driven by any single descriptor family: ten global physicochemical scalars recover 91% of full-feature performance. Finally, explicitly modelling label dependence yields targeted benefits for scarce activities and supports ranking which activity to assay next from partial positive evidence.

Machine Learning Prediction and Experimental Validation of Antimicrobial Peptide Activity Differences against Gram-Positive and Gram-Negative Bacteria

Antimicrobial peptides (AMPs) are primary candidates for addressing bacterial resistance. Although their target spectrum specificity varies significantly between Gram-positive and Gram-negative bacteria, current predictive models generally lack experimental validation. In this study, we constructed various machine learning models based on known sequences to systematically evaluate the performance of k-mer frequencies, physicochemical properties, and hybrid features in distinguishing the AMP target specificity. Results indicated that the random forest model based on eight key physicochemical properties performed best, achieving a test set accuracy of 82.09% with balanced classification and robust generalization. Feature importance analysis revealed that hydrophilicity and isoelectric point (pI) are the core physicochemical factors determining the target spectrum differences. The model was rigorously validated through a dual-track approach: first, via the synthesis and in vitro testing of 18 novel protozoan-derived AMPs (overall accuracy 66.67%) and, second, through a blind test on 55 independent external sequences, achieving a robust accuracy of 81.82%. Furthermore, the framework successfully identified candidates with potent activity against multidrug-resistant pathogens including Pseudomonas aeruginosa and Klebsiella pneumoniae. This experimentally validated predictive framework provides a reliable computational tool for the high-throughput screening and rational design of targeted antimicrobial peptides.

Transformer-based Discovery of Antimicrobial Peptides and Prediction of their Antibacterial Activity

With the worsening crisis of antimicrobial resistance, many researchers are now exploring new types of antibacterial agents, and antimicrobial peptides (AMPs) have begun to attract much attention. However, the experimental identification of AMPs in the large space of natural and synthetic sequences is still slow, expensive and labour-intensive. AMP-Transformer is a two-stage deep learning framework that combines self-supervised pre-training on large-scale unlabelled protein corpora with supervised fine-tuning on curated AMP datasets in this study. The model is a multi-layer bidirectional Transformer encoder that learns contextual residue representations via masked residue modelling, and is then fine-tuned for two coupled tasks: binary discrimination of AMPs from non-AMPs and regression of minimum inhibitory concentration (MIC) values. We used the benchmark datasets built by DBAASP v3, DRAMP 4.0 and other recently published experimental collections for evaluation. On an independent test set, AMP-Transformer had an accuracy of 95.3%, a Matthews correlation coefficient of 0.906, and an area under the receiver operating characteristic curve (AUC-ROC) of 0.986, and outperformed the support vector machine, random forest, convolutional, recurrent and hybrid baselines significantly. Ablation studies show that self-supervised pre-training and multi-head self-attention are the two main contributors, accounting for about 4% of the accuracy increase. Analysis of the learned attention maps shows that the model has independently learned the amphipathic periodicity and cationic residue enrichment characteristic of membrane-active peptides, thereby providing a degree of mechanistic interpretability that is rare among black-box predictors. A subsequent screening of metagenomic open reading frames also produced a ranked list of candidate AMPs with low sequence identity to any training examples, demonstrating the value of the framework for early-stage discovery. Based on the above results, Transformer-based protein language models are relatively stable, interpretable and scalable paradigms for AMP discovery and activity prediction, and they can help establish a practical computational pipeline that selects promising peptide candidates for experimental validation at a lower cost compared with traditional methods.

Integrating multi-species and multi-antibiotic resistance classification with MALDI-TOF: a deep learning approach to predict AMR

A novel multi-label, multi-bacteria classification framework (MLMBC) that simultaneously predicts AMR across multiple bacterial species and antibiotics using MALDI-TOF mass spectrometry data and Convolutional Neural Networks (CNNs).

Integrated Genome-Wide Association Study and Machine Learning Approach for Characterizing the Determinants of Biofilm Formation in Staphylococcus aureus.