Current forensic DNA phenotyping (FDP) has focused largely on pigmentation traits and has been developed and validated mainly in Euro-American populations, limiting its transferability to East Asian populations and leaving complex morphological traits underexplored. Body shape is a key component of appearance descriptions in forensic practice—frequently reported by eyewitnesses and visible in surveillance footage—yet standardized frameworks for body-shape evaluation and prediction remain limited. In Chinese populations, body shape shows substantial phenotypic diversity and may offer high practical utility for appearance inference, potentially complementing pigmentation traits that are often less variable in this context.

Here, we propose the oral microbiome as a non-invasive, information-rich biomarker to support body-shape inference, motivated by reports that oral microbial community profiles track host metabolic and inflammatory states and associate with adiposity-related phenotypes. We introduce a two-stage framework to infer perceived body shape in a Chinese cohort: Stage I uses oral microbiome features to independently predict six categorical objective indicators—body mass index (BMI), body fat percentage (BFP), waist-to-height ratio (WHtR), waist-to-hip ratio (WHR), waist circumference (WC), and visceral fat index (VFI)—and Stage II maps the resulting six-indicator profile to subjective body-shape perception.

We collected these six categorized indicators, three-view body photographs, and subjective body-shape assessments from 118 volunteers. Using these data, we built a calibration model linking objective indicator profiles to subjective body-shape categories, achieving good discriminative performance (AUC ≈ 0.90). This calibration provides a standardized basis for constructing a Chinese body-shape evaluation system and reference database, enabling consistent labeling for downstream predictive modeling and validation.

To support Stage I prediction, we performed 16S rRNA sequencing on paired oral samples from 90 volunteers. Saliva and tongue-coating microbiomes showed significant niche-specific differences (PERMANOVA R² = 0.177, P = 0.001), indicating that these sample types should be treated as distinct inputs and modeled separately. As a proof of concept for Stage I, machine-learning models trained on the salivary oral microbiome at the genus level achieved robust performance for BMI category classification (AUC = 0.926 for normal vs. overweight; 0.725 for overweight vs. obese). Future work with larger, age- and sex-stratified cohorts and broader coverage across the full spectrum of body types will help validate and extend the model’s generalizability across subgroups within Chinese populations and the tails of the body-shape distribution.

Authors

• Jiaxun Chen (West China School of Basic Medical Sciences & Forensic Medicine, Sichuan University, Chengdu, Sichuan 610041, P.R. China., China)
• Tianyu Li (West China School of Basic Medical Sciences & Forensic Medicine, Sichuan University, Chengdu, Sichuan 610041, P.R. China., China)
• Ye Duan (West China School of Basic Medical Sciences & Forensic Medicine, Sichuan University, Chengdu, Sichuan 610041, P.R. China., China)
• Xianliang Chen (West China School of Basic Medical Sciences & Forensic Medicine, Sichuan University, Chengdu, Sichuan 610041, P.R. China., China)
• Nuo Chen (West China School of Basic Medical Sciences & Forensic Medicine, Sichuan University, Chengdu, Sichuan 610041, P.R. China., China)
• Qian Wang (Laboratory of Metabolomics and Gynecological Disease Research, West China Second University Hospital, Sichuan University, Chengdu, Sichuan 610041, P. R. China, China)
• Zhilong Li (Laboratory of Molecular Translational Medicine, West China Second University Hospital, Sichuan University, Chengdu, Sichuan 610041, P. R. China, China)