A Scalable Biological Clock for Metabolic Disease Prediction from the Phenome India Cohort
- 1Department of Biology, Trivedi School of Biosciences, Ashoka University, Sonipat, Haryana 131029, India
Abstract
BackgroundIndia has a rising incidence of chronic non-communicable diseases, making it a major healthcare burden today. Growing evidence suggests that chronic low-grade inflammation links ageing with cardiometabolic disorders, captured by the emerging concept of inflammaging. However, most evidence on biological ageing comes from Western populations, with no similar models developed for the Indian population. Given the countrys distinctive genetic makeup, unique exposome, and heterogeneous NCD presentation, Western models may not capture inflammaging and its effects in the Indian population.
MethodsWe analysed baseline data from 4,240 adults in the Phenome India CSIR Health Cohort Knowledgebase (PI CheCK), a nationwide multi-centre cohort. Participants were stratified into eight cardiometabolic phenotype groups by BMI (Asian cut off), blood pressure and HbA1c status. We trained a Super Learner ensemble to predict chronological age in the lean normotensive-normoglycaemic reference group (n=615) using 44 plasma cytokines, sex, haemoglobin, and bioimpedance-derived visceral fat area, per cent body fat, and total body water. Performance was assessed by repeated five-fold cross-validation and in a held-out healthy test set. Calibrated biological age acceleration was then estimated in the remaining 3,625 participants.
ResultsMedian age was 51.0 years (IQR 41.0 to 62.0) and 49.4% were female. The Super Learner outperformed elastic net and XGBoost comparators. Permutation importance identified visceral fat area, per cent body fat, CTACK, SDF1a, haemoglobin and sex as leading contributors, with body composition measures accounting for the largest share, indicating an immune-metabolic rather than cytokine-only signal. Biological age acceleration was concentrated in overweight/obese phenotypes. Lean phenotypes showed acceleration close to the reference (0.32 0.50 years).
ConclusionsCytokine and body composition measures capture a quantifiable immunometabolic ageing signal in a South Asian cohort, with acceleration driven predominantly by adiposity. External validation and longitudinal follow up are required.
Related
- N-Ethylpentedrone: first hair detection and human metabolic profiling across multiple biological matrices using UHPLC-HRMS, molecular networking and in silico prediction
- The landscape of polygenic score prediction accuracy and transportability across the phenome in ancestrally diverse populations
- Prediction Models Versus Fibrosis-4 Index (FIB-4) for Advanced Fibrosis in Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD): A Verification-Audited Systematic Review and Meta-Analysis
- Design and Synthesis of New Thiazole Derivatives for Alzheimer’s Disease Therapy as a Potential Cholinesterase Inhibitor; Toxicity Prediction and Molecular Docking Studies and Biological Evaluation
- The natural history and individualized prediction of liver stiffness-based fibrosis risk in metabolic dysfunction-associated steatotic liver disease
BibTeX
@misc{tiwari2026scalable,
title = {{A Scalable Biological Clock for Metabolic Disease Prediction from the Phenome India Cohort}},
author = {P. Tiwari and M. Garg and S. Pattanayak and I. Sarkar and R. Roy and N. Bhatraju and A. Verma and S. R. K and S. Prakash and V. S. Kumar and M. A. Uddin and N. Rawat and A. Sahu and Y. Kumar and P. H. Leuva and A. Mridha and V. Yenamandra and A. P. Singh and A. Mishra and S. Raychaudhuri and K. B. Tallapaka and G. R. Chandak and M. J. Kulkarni and M. Dharne and R. Wahengbam and J. Kalita and P. Manna and U. Subudhi and S. Majumder and P. Chakraborty and K. Chaudhary and S. Sengupta and Phenome India Consortium and V. Sardana and S. Chatterjee and D. Ganguly},
howpublished = {medRxiv},
year = {2026},
month = aug,
doi = {10.64898/2026.08.29.26361656},
url = {https://www.medrxiv.org/content/10.64898/2026.08.29.26361656v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - A Scalable Biological Clock for Metabolic Disease Prediction from the Phenome India Cohort
AU - P. Tiwari
AU - M. Garg
AU - S. Pattanayak
AU - I. Sarkar
AU - R. Roy
AU - N. Bhatraju
AU - A. Verma
AU - S. R. K
AU - S. Prakash
AU - V. S. Kumar
AU - M. A. Uddin
AU - N. Rawat
AU - A. Sahu
AU - Y. Kumar
AU - P. H. Leuva
AU - A. Mridha
AU - V. Yenamandra
AU - A. P. Singh
AU - A. Mishra
AU - S. Raychaudhuri
AU - K. B. Tallapaka
AU - G. R. Chandak
AU - M. J. Kulkarni
AU - M. Dharne
AU - R. Wahengbam
AU - J. Kalita
AU - P. Manna
AU - U. Subudhi
AU - S. Majumder
AU - P. Chakraborty
AU - K. Chaudhary
AU - S. Sengupta
AU - Phenome India Consortium
AU - V. Sardana
AU - S. Chatterjee
AU - D. Ganguly
PY - 2026
DA - 2026/08/29
PB - medRxiv
DO - 10.64898/2026.08.29.26361656
UR - https://www.medrxiv.org/content/10.64898/2026.08.29.26361656v1
AB - BackgroundIndia has a rising incidence of chronic non-communicable diseases, making it a major healthcare burden today. Growing evidence suggests that chronic low-grade inflammation links ageing with cardiometabolic disorders, captured by the emerging concept of inflammaging. However, most evidence on biological ageing comes from Western populations, with no similar models developed for the Indian population. Given the countrys distinctive genetic makeup, unique exposome, and heterogeneous NCD presentation, Western models may not capture inflammaging and its effects in the Indian population. MethodsWe analysed baseline data from 4,240 adults in the Phenome India CSIR Health Cohort Knowledgebase (PI CheCK), a nationwide multi-centre cohort. Participants were stratified into eight cardiometabolic phenotype groups by BMI (Asian cut off), blood pressure and HbA1c status. We trained a Super Learner ensemble to predict chronological age in the lean normotensive-normoglycaemic reference group (n=615) using 44 plasma cytokines, sex, haemoglobin, and bioimpedance-derived visceral fat area, per cent body fat, and total body water. Performance was assessed by repeated five-fold cross-validation and in a held-out healthy test set. Calibrated biological age acceleration was then estimated in the remaining 3,625 participants. ResultsMedian age was 51.0 years (IQR 41.0 to 62.0) and 49.4% were female. The Super Learner outperformed elastic net and XGBoost comparators. Permutation importance identified visceral fat area, per cent body fat, CTACK, SDF1a, haemoglobin and sex as leading contributors, with body composition measures accounting for the largest share, indicating an immune-metabolic rather than cytokine-only signal. Biological age acceleration was concentrated in overweight/obese phenotypes. Lean phenotypes showed acceleration close to the reference (0.32 0.50 years). ConclusionsCytokine and body composition measures capture a quantifiable immunometabolic ageing signal in a South Asian cohort, with acceleration driven predominantly by adiposity. External validation and longitudinal follow up are required.
ER -