Fig. 6

Overview of machine-learning pipeline implementing biological variables across a spectrum of gathered information. From the patient population undergoing coronary artery bypass graft surgery (CABG), physiological parameters (demographics, health reports, etc.) and atrial tissue were used for subsequent analyses. From cardiac tissue genomic (mitochondrial DNA), epigenomic (TFAM promoter CpG methylation), and biochemical (nuclear and mitochondrial function) were assessed. Cumulatively, the biological data was processed through tree ensembles in SHAP and validated through CART analysis with tenfold cross validation. Using these machine-learning algorithms, graphical depictions and biomarker feature importance are able to be derived, allowing for prediction of the onset and progression of diabetes. Ultimately, by using biological data at the genomic and epigenomic level, it allows for precision medicine approaches and more personalized diagnostics and prognostics. TFAM: transcription factor A, mitochondrial; mtDNA: mitochondrial DNA; CpG: cytosine nucleotide followed by a guanine nucleotide; CART: Classification and Regression Trees; SHAP: SHapley Additive exPlanations