Machine learning approach on plasma proteomics identifies signatures associated with obesity in the KORA FF4 cohort
Authors' conclusion
The protein signatures identified in our study showed promising predictive potential for obesity. These findings warrant further investigation to evaluate their clinical applicability, offering insights into obesity prevention and treatment strategies.
Study context
- Population
- Individuals in the KORA FF4 study; 809 plasma proteins measured
- Participants
- 2,045 individuals
- Exposure
- Obese vs non-obese
- Obesen = 564
- Non-obesen = 1,481reference
- Outcome
- Obesity-associated plasma proteins and ML prediction of obesity
Methodological detailsmachine learning, Priority-Lasso feature selection, Mendelian randomisation
- Machine learning
- Priority-Lasso feature selection
- Mendelian randomisation
Main finding
As reported in the abstract
- Proteins significantly associated with obesity (basic model logistic regression and Priority-Lasso)
- 16
- AUC, SVM
- 0.820
- AUC, RF
- 0.805
- AUC, KNN
- 0.791
- AUC, Adaboost
- 0.819
Observational data: associations, not effects.
Further findings
Enrichment analyses
- Immune response, lipid metabolism and inflammation regulation linked to obesity
Correlation with risk factors
- All 16 proteins correlated with obesity-related risk factors such as blood pressure and lipid levels
Mendelian randomisation
- AFM, CRP and CFH identified as causal and potentially modifiable proteins
Original publication
Machine learning approach on plasma proteomics identifies signatures associated with obesity in the KORA FF4 cohort. Diabetes Obes Metab. Published 27 February 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →