Forensic prediction of sex, age, height, body mass index, hip-to-waist ratio, smoking status and lipid lowering drugs using epigenetic markers and plasma proteins.

Llobet MO, Johansson Å, Gyllensten U, Allen M, Enroth S

Forensic Sci Int Genet 65 (-) 102871 [2023-07-00; online 2023-04-07]

The prediction of human characteristics from blood using molecular markers would be very helpful in forensic science. Such information can be particularly important in providing investigative leads in police casework from, for example, blood found at crime scenes in cases without a suspect. Here, we investigated the possibilities and limitations of predicting seven phenotypic traits (sex, age, height, body mass index [BMI], hip-to-waist [WTH] ratio, smoking status and lipid-lowering drug use) using either DNA methylation or plasma proteins separately or in combination. We developed a prediction pipeline starting with the prediction of sex followed by sex-specific, stepwise, individual age, sex-specific anthropometric traits and, finally, lifestyle-related traits. Our data revealed that age, sex and smoking status can be accurately predicted from DNA methylation alone, while the use of plasma proteins was highly accurate for prediction of the WTH ratio, and a combined analysis of the best predictions for BMI and lipid-lowering drug use. In unseen individuals, age was predicted with a standard error of 3.3 years for women and 6.5 years for men, while the accuracy in smoking prediction across both men and women was 0.86. In conclusion, we have developed a stepwise approach for the de-novo prediction of individual characteristics from plasma proteins and DNA methylation markers. These models are accurate and may provide valuable information and investigative leads in future forensic casework.

Bioinformatics Support for Computational Resources [Service]

NGI SNP genotyping [Service]

NGI Uppsala (SNP&SEQ Technology Platform) [Service]

National Genomics Infrastructure [Service]

PubMed 37054667

DOI 10.1016/j.fsigen.2023.102871

Crossref 10.1016/j.fsigen.2023.102871

pii: S1872-4973(23)00046-7


Publications 9.5.0