Weighted spectrochemical correlation network analysis-guided GA-PLSR: A potential spectral “fluid biopsy” approach for quantitative assessment of cardiac metabolites in diabetic cardiomyopathy

Abstract

There is a vast body of research indicating that pronounced metabolic abnormalities are important indications of early-stage diabetic cardiomyopathy (DbCM), which is hardly detected clinically. Analytical tools that could non-invasively monitor the processes of these DbCM-related metabolites in the myocardium are highly desired. Here, we present a spectral “fluid biopsy” strategy to non-invasively predict the cardiometabolic spectral markers (CMSMs) in the myocardium of db/db mice. We employed ATR-FTIR spectroscopy to capture the spectrochemical profiles of the plasma, urine, saliva, and myocardium of db/db mice ranging in age from 7 to 21 weeks. Then, we established a formal framework called weighted spectrochemical correlation network analysis (WSCNA) to identify specific biofluid spectral modules that were highly correlated with CMSMs. Thereafter, we regressed CMSMs on the corresponding biofluid spectral modules substantially associated with CMSMs using the GA-PLSR algorithm. Our results demonstrated that the regression models based on biofluid spectral modules selected by WSCNA had a good ability to predict CMSMs involving diabetic myocardial lipid metabolites (e.g., total lipids, longer-chain fatty acids, lipid carbonyl esters, and unsaturated lipids) with an average cross-validated R2 value of 0.84 and an average ratio of performance to deviation of 2.68. This study demonstrates the potential of the WSCNA-GA-PLSR technique for developing biofluid infrared spectro-diagnostic models for non-invasive assessment of DbCM-related CMSMs.