Body fat distribution can substantially affect appearance and therefore estimating body mass index (BMI) may aid in reconstructing the physical phenotype of an unidentified individual. Importantly, BMI is recognized to have a strong genetic component and also stems from lifestyle influences. Epigenetic markers have been successfully employed in BMI prediction. However, current models require hundreds of CpG sites, which may limit their utility in forensic settings. Moreover, most of the existing predictors rely on blood-derived data, overlooking other forensically relevant tissues. In this study, we assessed the potential to predict BMI by leveraging both genome-wide DNA methylation and SNP data generated using microarray technology across various tissue types: blood, buccal cells, and semen. The most informative markers were selected to optimize the balance between predictive accuracy and forensic applicability.

In a sample of 624 adult individuals, the top ten CpG sites accounted for nearly twice the variance in BMI compared to the top ten SNPs. The polygenic risk score based on over 900 SNPs showed a highly significant, though low-to-moderate, correlation with BMI (r=0.25, p=2.7×10⁻¹⁰). Elastic Net regression applied to blood-derived data yielded 30 CpG predictors with non-zero coefficients, favoring a simpler, more regularized model while maintaining a low prediction error. When projected onto an independent validation set (N=112), the model achieved a mean absolute error (MAE) of 2.9 kg/m², which is comparable to values reported for other large-scale predictors. Notably, an alternative binomial regression model achieved an AUC of 0.74, showing high sensitivity (82.1%) but lower specificity (62.2%) for identifying individuals with overweight status. 
The developed predictors exhibited significantly higher prediction error in buccal swabs (N=227) and in semen (N=165); therefore, tissue-specific remodeling was applied, resulting in cross-validated MAE of 3.2 and 2.5 kg/m², respectively. Finally, the BMI models were applied to twin samples with available 3D facial scans (N=15 pairs) to further assess model sensitivity and evaluate whether epigenetic BMI can serve as a meaningful covariate in facial reconstruction. 

Our study demonstrates that BMI can be predicted from DNA across different tissues using a modest number of markers, with the compact model supporting its forensic applicability. In conclusion, BMI models may enhance forensic DNA phenotyping including aiding facial reconstruction.

The study received funding from the European Union’s Horizon Europe Programme under Grant agreement no. 101225631 and from the National Centre for Research and Development in Poland (no. DOBBIO10/06/01/2019).

Authors

• Ewelina Pośpiech (Department of Genomics and Forensic Genetics, Pomeranian Medical University in Szczecin, Poland)
• Kamila Marszałek (Department of Genomics and Forensic Genetics, Pomeranian Medical University in Szczecin, Poland)
• Maria Wróbel (Institute of Forensic Research in Krakow, Poland)
• Julia Zacharczuk (Department of Genomics and Forensic Genetics, Pomeranian Medical University in Szczecin, Poland)
• Balakrishnan Subramanian (Institute of Zoology and Biomedical Research, Jagiellonian University, Poland)
• Aleksandra Pisarek-Pacek (Institute of Zoology and Biomedical Research, Jagiellonian University, Poland)
• Antonia Heidegger (Institute of Legal Medicine, Medical University of Innsbruck, Austria)
• Miguel Boullón-Cassau (Forensic Genetics Unit, Institute of Forensic Sciences, University of Santiago de Compostela, Spain)
• Adrián Ambroa-Conde (Forensic Genetics Unit, Institute of Forensic Sciences, University of Santiago de Compostela, Spain)
• Mafalda Silva (Forensic Genetics Unit, Institute of Forensic Sciences, University of Santiago de Compostela, Spain)
• Katarzyna Urbanelis (Institute of Zoology and Biomedical Research, Jagiellonian University, Poland)
• Bożena Wysocka (Central Forensic Laboratory of the Police, Warsaw, Poland, Poland)
• Aneta Sitek (Department of Anthropology, University of Lodz, Poland)
• Magdalena Spólnicka (Center for Forensic Science University of Warsaw, Poland)
• Andrzej Ossowski (Department of Genomics and Forensic Genetics, Pomeranian Medical University in Szczecin, Poland)
• Walther Parson (Institute of Legal Medicine, Medical University of Innsbruck, Austria)
• María Victoria Lareu (Forensic Genetics Unit, Institute of Forensic Sciences, University of Santiago de Compostela, Spain)
• Ana Freire-Aradas (Forensic Genetics Unit, Institute of Forensic Sciences, University of Santiago de Compostela, Spain)
• Wojciech Branicki (Institute of Zoology and Biomedical Research, Jagiellonian University, Poland)