Selective taste impairment in relapsing-remitting Multiple Sclerosis and Supervised Learning - driven diagnostic classification.

Taste disturbances have been largely overlooked in Multiple Sclerosis (MS), despite mounting evidence that neuroinflammation and demyelination may disrupt chemosensory pathways. This study investigated taste perception in People with MS (PwMS) compared with healthy controls (HCs), explored whether specific taste-related features could classify MS using Supervised Learning (SL), and examined associations with taste receptor gene polymorphisms. Using validated taste-strip tests, we found that PwMS exhibit marked impairments in bitter, salty, and sour identification, accompanied by a selective increase in bitter detection threshold, while sweet perception remained relatively preserved. Principal component analysis showed substantial overlap between groups, yet the SL CatBoost classifier achieved high accuracy (training: 89%; testing: 86%), indicating that multivariate taste-sensitivity patterns reliably distinguish PwMS from HCs. The SHAP algorithm revealed that bitter and citric-acid perception thresholds, along with correct identification scores, were the most important features for MS prediction, and sucrose-related features were also key factors. Although a TAS1R2 rs35874116 polymorphism differed in frequency between groups, its contribution to SL classification was minimal compared to behavioral taste measures. Overall, our findings, by showing specific and quantifiable taste patterns, support the integration of taste tests into the broader assessment of MS, with potential applications in early diagnosis, disease monitoring, and improving the quality of life of PwMS.