Machine learning approach to assess the association between anthropometric, metabolic, and nutritional status and semen parameters

Guillaume Bachelot; Antonin Lamaziere; Sebastien Czernichow; Celine Faure; Chrystelle Racine; Rachel Levy; Charlotte Dupont; Nutrition and Fertility (ALIFERT) Group

doi:10.4103/aja20247

Machine learning approach to assess the association between anthropometric, metabolic, and nutritional status and semen parameters

Asian J Androl. 2024 Apr 12. doi: 10.4103/aja20247. Online ahead of print.

Authors

Affiliations

¹ Sorbonne University School of Medicine, Saint-Antoine Research Center, INSERM UMR 938, 27 rue Chaligny, Paris 75012, France.
² Reproductive Biology Department-CECOS, Tenon Hospital, AP-HP.Sorbonne University, Paris 75020, France.
³ Clinical Metabolomics Department, Saint-Antoine Hospital, AP-HP, 27 rue Chaligny, Paris 75012, France.
⁴ Nutrition Department, Obesity Specialist Centre, Georges Pompidou European Hospital, AP-HP, Paris 75015, France.

PMID: 38624205
DOI: 10.4103/aja20247

Abstract

Many lifestyle factors, such as nutritional imbalance leading to obesity, metabolic disorders, and nutritional deficiency, have been identified as potential risk factors for male infertility. The aim of this study was to evaluate the relationship between semen parameters and anthropometric, metabolic and nutritional parameters. Relationship was first assessed individually, then after the application of a previously constructed and validated machine learning score that allows their combination. Anthropometric, metabolic, antioxidant, micronutrient, and sperm parameters from 75 men suffering from idiopathic infertility from four infertility centers in France (Jean-Verdier ART Center Hospital, Bondy; North Hospital ART Center, Saint-Étienne; Navarre Polyclinic ART Center, Pau; and Cochin Hospital ART Center, Paris) between September 2009 and December 2013 were collected. After assessing standard correlation analysis, a previously built machine learning model, providing a score ranging from 0 (the poorest) to 1 (the most favorable), was calculated for each man in the study cohort. This machine learning model, which separates infertile/fertile men with unexplained infertility on the basis of their bioclinical signature, provides a more holistic evaluation of the influence of the considered markers (anthropometric, metabolic, and oxidative status). We observed a significant correlation of some anthropometric, metabolic, and nutritional disorders with some sperm characteristics. Moreover, an unfavorable machine learning score was associated with a high level of sperm DNA fragmentation. Favorable anthropometric, metabolic, and oxidative patterns, which may reflect an appropriate lifestyle, appear to positively impact overall health, in particular reproductive function. This study, consistent with previous publications, suggests that beyond semen quality parameters, in an essential assessment of male fertility, other key factors should be taken into account. In this regard, the application of emerging artificial intelligence techniques may provide a unique opportunity to integrate all these parameters and deliver personalized care.