Plasma lipid metabolites as potential biomarkers for identifying individuals at risk of obesity-induced metabolic complications
Paula Emília Nunes Ribeiro Bellot, Erik Sobrinho Braga, Folorunsho Bright Omage, Francisca Leide da Silva Nunes, Severina Carla Vieira Cunha Lima, Clélia de Oliveira Lyra, Dirce Maria Lobo Marchioni, Lúcia Fátima Campos Pedrosa, Fernando Barbosa, Ljubica Tasić, Karine Cavalcanti Maurício Sena‐Evangelista
Abstract
Abstract Lipidomics studies have indicated an association between obesity and lipid metabolism dysfunction. This study aimed to evaluate and compare cardiometabolic risk factors, and the lipidomic profile in adults and older people. A cross-sectional study was conducted with 72 individuals, divided into two sex and age-matched groups: obese (body mass index—BMI ≥ 30 kg/m 2 ; n = 36) and non-obese (BMI < 30 kg/m 2 ; n = 36). The lipidomic profiles were evaluated in plasma using 1 H nuclear magnetic resonance ( 1 H-NMR) spectroscopy. Obese individuals had higher waist circumference (p < 0.001), visceral adiposity index (p = 0.029), homeostatic model assessment insulin resistance (HOMA-IR) (p = 0.010), and triacylglycerols (TAG) levels (p = 0.018). 1 H-NMR analysis identified higher amounts of saturated lipid metabolite fragments, lower levels of unsaturated lipids, and some phosphatidylcholine species in the obese group. Two powerful machine learning (ML) models— k -nearest neighbors ( k NN) and XGBoost (XGB) were employed to characterize the lipidomic profile of obese individuals. The results revealed metabolic alterations associated with obesity in the NMR signals. The models achieved high accuracy of 86% and 81%, respectively. The feature importance analysis identified signal at 1.50–1.60 ppm (–CO–CH 2 –CH 2 –, Cholesterol and fatty acid in TAG, Phospholipids) to have the highest importance in the two models.