In American journal of obstetrics & gynecology MFM
BACKGROUND : Hypertensive disorders of pregnancy(HDP) account for 3-10% of maternal-fetal morbidity and mortality worldwide and it is one among the leading causes of maternal deaths in developing countries like India.
OBJECTIVE : We aim to discover HDP-specific candidate urine metabolites as markers for HDP by applying integrated metabolomics and machine learning(ML) approaches.
STUDY DESIGN : The targeted urinary metabolomics study was conducted in 70 healthy pregnant controls and 133 pregnant patients having hypertension as cases. Hypertensive disorders of pregnancy specific metabolites for disease prediction were further extracted by univariate and multivariate statistical analysis. For machine learning analysis, 80% of data was used for training(79-HDP,42-HP), validation(27-HDP,14-HP) and 20% of data for test sets(27-HDP, 14-HP).
RESULTS : The statistical analysis by unpaired t-test revealed forty-four differential metabolites. Pathway analysis showed mainly that purine and thiamine metabolism were altered in HDP group compared to HP group. The area under ROC curve of the five most predominant metabolites was adenosine(0.98), adenosine monophosphate(AMP, 0.92), deoxyadenosine(0.89), thiamine(0.81), and thiamine monophosphate(TMP, 0.81). The best prediction accuracies were obtained by two machine learning(ML) models such as 95% for gradient boost(GBM), and 98% for decision tree(DT) among five used models. The ML models showed higher predictive performance for three metabolites i.e., TMP, AMP, and thiamine, among five metabolites. The combined accuracy of adenosine from all the models is 98.6 in training set and 95.6 in test set. Moreover, the predictive performance of adenosine was higher than other metabolites. The relative feature importance of adenosine was also observed in DT and GBM.
CONCLUSION : The adenosine and thiamine metabolites were found to differentiate HDP with HP subjects among other metabolites; hence these metabolites can serve as a promising non-invasive marker for HDP detection.
Varghese Bincy, Jala Aishwarya, Meka Soumya, Adla Deepthi, Jangili Shraddha, Talukdar R K, Mutheneni Srinivasa Rao, Borkar Roshan M, Adela Ramu
2022-Dec-01
Adenosine, Biomarkers, Hypertension disorders of pregnancy, Machine learning, Mass spectrometry, Metabolomics, Purine, Thiamine
