Animal Microbiome, Год журнала: 2025, Номер 7(1)

Опубликована: Апрель 24, 2025

The gut microbiome plays a crucial role in health and disease, influencing digestion, metabolism, immune function. Traditional analysis methods are often expensive, time-consuming, require specialized expertise, limiting their practical application clinical settings. Evolving artificial intelligence (AI) technologies present opportunities for developing alternative methods. However, the lack of transparency these limits ability clinicians to incorporate AI-driven diagnostic tools into healthcare systems. aim this study was investigate an AI approach that rapidly predicts different bacterial genera groups, specifically butyrate producers, from digital images fecal smears rhesus macaques (Macaca mulatta). In addition, improve transparency, we employed explainability uncover image features model's predictions. By integrating data with corresponding metagenomic sequencing information, deep learning (DL) machine (ML) algorithms successfully predicted 16 individual (area under curve (AUC) > 0.7) among 50 most abundant model successful predicting functional major producers (AUC 0.75) mixed group including fermenters short-chain fatty acid (SCFA) 0.81). For both models fermenters, experiments revealed no decline AUC when random noise added images. Increased blurring led gradual AUC. performance robust against impact shape smearing, stable maintained until patch 4 all as assessed through scrambling. No significant correlation detected between prediction probabilities total weight used smear; r = 0.30 ± 0.3 (p 0.1) 0.04 0.36 0.8) respectively. Our demonstrated predict wide range clinically relevant microbial groups based on smear. proved be smearing method, amount matter. This introduces rapid, non-invasive, cost-effective method profiling, potential applications veterinary diagnostics.

Язык: Английский