Abstract

Personalized medicine relies on successful identification of genome-wide variations that governs inter-individual differences in phenotypes and system level outcomes. In Ayurveda, assessment of composite constitution types "Prakriti" forms the basis for risk stratification, predicting health and disease trajectories and personalized recommendations. Here, we report a novel method for identifying pleiotropic genes and variants that associate with healthy individuals of three extreme and contrasting "Prakriti" constitutions through exome sequencing and state-of-the-art computational methods. Exome Seq of three extreme Prakriti types from 108 healthy individuals 54 each from genetically homogeneous populations of North India (NI, Discovery cohort) and Western India (VADU, Replication cohort) were evaluated. Fishers Exact Test was applied between Prakriti types in both cohorts and further permutation based p-value was used for selection of exonic variants. To investigate the effect of sample size per genetic association test, we performed power analysis. Functional impact of differentiating genes and variations were inferred using diverse resources -Toppfun, GTEx, GWAS, PheWAS, UK Biobank and mouse knockdown/knockout phenotype (MGI). We also applied supervised machine learning approach to evaluate the association of exonic variants with multisystem phenotypes of Prakriti. Our targeted investigation into exome sequencing from NI (discovery) and VADU (validation) cohorts datasets provide ~7,000 differentiating SNPs. Closer inspection further identified a subset of SNPs (2407 (NI) and 2393 (VADU)), that mapped to an overlapping set of 1181 genes. This set can robustly stratify the Prakriti groups into three distinct clusters with distinct gene ontological (GO) enrichments. Functional analysis further strengthens the potential pleiotropic effects of these differentiating genes/variants and multisystem phenotypic consequences. Replicated SNPs map to some very prominent genes like FIG4, EDNRA, ANKLE1, BCKDHA, ATP5SL, EXOCS5, IFIT5, ZNF502, PNPLA3 and IL6R. Lastly, multivariate analysis using random forest uncovered rs7244213 within urea transporter SLC14A2, that associate with an ensemble of features linked to distinct constitutions. Our results reinforce the concept of integration of Prakriti based deep phenotypes for risk stratification of healthy individuals and provides markers for early actionable interventions.