Integrating Machine Learning with MALDI-TOF Mass Spectrometry for Rapid and Accurate Antimicrobial Resistance Detection in Clinical Pathogens

International Journal of Molecular Sciences, Journal Year: 2025, Volume and Issue: 26(3), P. 1140 - 1140

Published: Jan. 28, 2025

Antimicrobial resistance (AMR) is one of the most pressing public health challenges 21st century. This study aims to evaluate efficacy mass spectral data generated by VITEK® MS instruments for predicting antibiotic in Staphylococcus aureus, Escherichia coli, and Klebsiella pneumoniae using machine learning algorithms. Additionally, potential pre-trained models was assessed through transfer analysis. A dataset comprising 2229 spectra collected, classification algorithms, including Support Vector Machines, Random Forest, Logistic Regression, CatBoost, were applied predict resistance. CatBoost demonstrated a clear advantage over other models, effectively handling complex non-linear relationships within achieving an AUROC 0.91 F1 score 0.78 E. coli. In contrast, yielded suboptimal results. These findings highlight gradient-boosting techniques enhance prediction, particularly with from less conventional platforms like MS. Furthermore, identification specific biomarkers SHAP values indicates promising clinical applications early diagnosis. Future efforts focused on standardizing refining algorithms could expand utility these approaches across diverse environments, supporting global fight against AMR.

Language: Английский

---

The future of pharmaceuticals: Artificial intelligence in drug discovery and development

Journal of Pharmaceutical Analysis, Journal Year: 2025, Volume and Issue: unknown, P. 101248 - 101248

Published: Feb. 1, 2025

Language: Английский

---

Applications of Artificial Intelligence in Drug Repurposing

Advanced Science, Journal Year: 2025, Volume and Issue: unknown

Published: March 6, 2025

Drug repurposing identifies new therapeutic uses for the existing drugs originally developed different indications, aiming at capitalizing on established safety and efficacy profiles of known drugs. Thus, it is beneficial to bypass early stages drug development, reduction time cost associated with bringing therapies market. Traditional experimental methods are often time-consuming expensive, making artificial intelligence (AI) a promising alternative due its lower cost, computational advantages, ability uncover hidden patterns. This review focuses availability AI algorithms in their positive specific roles revealing drugs, especially being integrated virtual screening. It shown that excel analyzing large-scale datasets, identifying complicated patterns responses from these predictions potential repurposing. Building insights, challenges remain developing efficient future research, including integrating drug-related data across databases better repurposing, enhancing efficiency, advancing personalized medicine.

Language: Английский

---

Pre-trained molecular representations enable antimicrobial discovery

Nature Communications, Journal Year: 2025, Volume and Issue: 16(1)

Published: April 10, 2025

The rise in antimicrobial resistance poses a worldwide threat, reducing the efficacy of common antibiotics. Determining activity new chemical compounds through experimental methods remains time-consuming and costly. While compound-centric deep learning models promise to accelerate this search prioritization process, current strategies require large amounts custom training data. Here, we introduce lightweight computational strategy for discovery that builds on MolE (Molecular representation redundancy reduced Embedding), self-supervised framework leverages unlabeled structures learn task-independent molecular representations. By combining with available, experimentally validated compound-bacteria data, design general predictive model enables assessing respect their potential. Our correctly identifies recent growth-inhibitory are structurally distinct from Using approach, discover de novo, confirm, three human-targeted drugs as growth inhibitors Staphylococcus aureus. This offers viable, cost-effective antibiotic discovery.

Language: Английский

---

Databases of ligand-binding pockets and protein-ligand interactions

Computational and Structural Biotechnology Journal, Journal Year: 2024, Volume and Issue: 23, P. 1320 - 1338

Published: March 24, 2024

Many research groups and institutions have created a variety of databases curating experimental predicted data related to protein-ligand binding. The landscape available is dynamic, with new emerging established becoming defunct. Here, we review the current state that contain binding pockets interactions. We compiled list such databases, fifty-three which are currently for use. discuss variation in how defined summarize pocket-finding methods. organize into subgroups based on goals contents, describe standard use cases. also illustrate within same protein characterized differently across different databases. Finally, assess critical issues sustainability, accessibility redundancy.

Language: Английский

---