GENOTYPE-BASED MODEL FOR PREDICTING OVERWEIGHT AND OBESITY IN THE PERUVIAN POPULATION USING MACHINE LEARNING: A PILOT STUDY

López A.G.; García-Quispes W.A.; Quiroz D.L. · Revista de la Facultad de Medicina Humana (2026) · DOI: 10.25176/rfmh.v26i3.8471

Resumen

Introduction: Genetic prediction of obesity using machine learning (ML) has not been evaluated in small, imbalanced Peruvian samples. Objective: To assess the feasibility and limitations of ML models based on 11 metabolic single-nucleotide polymorphisms (SNPs), explore the contribution of age and sex in a secondary random forest (RF) analysis, describe allele frequencies, and estimate the sample size required for future studies. Methods: Forty-eight adults were genotyped. Logistic regression, RF, conditional inference tree, and k-nearest neighbors models were evaluated using leave-one-out cross-validation. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), balanced accuracy, Matthews correlation coefficient (MCC), F1-score, and confusion matrix. Results: All SNPs were in Hardy-Weinberg equilibrium, and none showed a significant univariate association with overweight/obesity (all p > 0.050), consistent with statistical power limited to detecting odds ratios (ORs) ≥3.50. Allele frequencies for all 11 variants were documented. The RF achieved the highest reported AUC after orientation correction (0.735), but failed to correctly classify any normal-weight participant. Its specificity was 0.000, balanced accuracy was 0.456, and MCC was -0.166. None of the models outperformed the balanced performance expected by chance. Post hoc power analysis estimated that approximately 200 participants would be required to detect an OR of 1.50 and 300 participants to detect an OR of 1.30. Conclusions: None of the evaluated models was clinically useful. In small, imbalanced datasets, the AUC, when interpreted in isolation, may overestimate model performance and should be complemented with robust performance metrics. © Artículo publicado por la Revista de la Facultad de Medicina Humana de la Universidad Ricardo Palma. Es un artículo de acceso abierto, distribuido bajo los términos de la Licencia Creative Commons: Creative Commons Attribution 4.0 International, CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), que permite el uso no comercial, distribución y reproducción en cualquier medio, siempre que la obra original sea debidamente citada. Para uso comercial, por favor póngase en contacto con revista.medicina@urp.pe.

Instituciones: Universidad Nacional Mayor de San Marcos