medRxiv preprint doi: https://doi.org/10.1101/2021.07.05.21260032; this version posted July 7, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC 4.0 International license . Comparing the genetic and environmental architecture of blood count, blood biochemistry and urine biochemistry biological ages with machine learning Alan Le Goallec1,2+, Samuel Diai1+, Théo Vincent1, Chirag J. Patel1* 1Department of Biomedical Informatics, Harvard Medical School, Boston, MA, 02115, USA 2Department of Systems, Synthetic and Quantitative Biology, Harvard University, Cambridge, MA, 02118, USA +Co-first authors *Corresponding author Contact information: Chirag J Patel
[email protected] Abstract While a large number of biological age predictors have been built from blood samples, a blood count-based biological age predictor is lacking, and the genetic and environmental factors associated with blood-measured accelerated aging remain elusive. In the following, we leveraged 31 blood count biomarkers measured from 489,079 blood samples, 28 blood biochemistry biomarkers measured from 245,147 blood samples, and four urine biochemistry biomarkers measured from 158,381 samples to build three distinct biological age predictors by training machine learning models to predict age. Blood biochemistry significantly outperformed blood count and urine biochemistry in terms of age prediction (RMSE: 5.92+-0.02 vs. 7.60+-0.02 years and 7.72+-0.04 years). We performed genome wide association studies [GWASs], and found accelerated blood biochemistry, blood count and urine biochemistry aging to be respectively 26.2+-0.3%, 18.1+-0.2% and 10.5±0.5% GWAS-heritable.