28 июля 2025
Gut Microbiome Signatures in Multiple Sclerosis: A Case-Control Study with Machine Learning and Global Data Integration
Margarita V. Neklesova, Karine S. Sogomonyan, Ivan A. Golovkin, Nikolay I. Shirokiy, Sofia O. Vershinina, Sofia A. Tsvetikova, Julia E. Korzhova, Mariya N. Zakharova and Elena V. Gnedovskaya
Biomedicines 2025, 13, 1806
Background/Objectives: Gut dysbiosis has been implicated in multiple sclerosis (MS), but
microbial signatures remain inconsistent across studies. Machine learning (ML) algorithms
based on global microbiome data integration can reveal key disease-associated microbial
biomarkers and new insights into MS pathogenesis. This study aimed to investigate gut
microbial signatures associated with MS and to evaluate the potential of ML for diagnostic
applications. Methods: Fecal samples from 29 relapsing–remitting MS patients during
exacerbation and 27 healthy controls were analyzed using 16S rRNA gene sequencing.
Differential abundance analysis was performed, and data were integrated with 29 published
studies. Four ML models were developed to distinguish MS-associated microbiome profiles.
Results: MS patients exhibited reduced levels of Eubacteriales (p = 0.037), Lachnospirales
(p = 0.021), Oscillospiraceae (p = 0.013), Lachnospiraceae (p = 0.012), Parasutterella (p = 0.018),
Faecalibacterium (p = 0.004), and higher abundance of Lachnospiraceae UCG-008 (p = 0.045)
compared to healthy controls. The Light Gradient Boosting Machine classifier demonstrated
the highest performance (accuracy: 0.88, AUC-ROC: 0.95) in distinguishing MS microbiome
profiles from healthy controls. Conclusions: This study highlights specific microbiome
dysbiosis in MS patients and supports the potential of ML for diagnostic applications.
Further research is needed to elucidate the mechanistic role of these microbial alterations in
MS progression and their therapeutic utility.