Introduction

Forensic DNA Phenotyping (FDP) requires robust models that distinguish between biological markers and social identity. In highly admixed populations like Brazil, identifying the specific determinants of these traits is essential for improving the accuracy of biological profiling and actionable investigative leads.

Methods

We evaluated 611 individuals from the São Paulo metropolitan area. Constitutive melanin density (M-index) was measured via reflectance spectrophotometry at a sun-protected site. We tested associations of skin color genetic predictions obtained using the HIrisPlex-S system, 74 individual pigmentation SNPs (including the 41 present in HIrisPlex-S), global genetic ancestry, and sociodemographic data. We employed regression models to determine the variance explained by each category of predictors for both the M-index and self-reported skin color.

Results

The M-index was strongly associated with HIrisPlex-S "Dark" probabilities and African genetic ancestry, which together explained ~6% of its variance. However, a model using only individual SNPs in the genes SLC45A2, MC1R, IRF4, BNC2, HHIP, explained 12.7% of melanin variability, with rs16891982 (SLC45A2) emerging as the strongest single predictor. 
Regarding self-reported identity, the predictors varied across categories:

• "Branco" (White): Self-identification was positively associated with European genetic ancestry, HIrisPlex-S predicted "Pale" and "Intermediate" skin, and younger age, while being inversely associated with the M-index.
• "Preto" (Black): This category was strongly predicted by African genetic ancestry and genetically predicted "Dark" and "Dark-to-Black" skin, alongside male sex. M-index, age and income did not show evidence of association with "Preto" identification.
• "Pardo" (Brown): This intermediate classification was positively associated with the M-index and genetically predicted "Dark" skin, but was uniquely and inversely associated with income.

Furthermore, several individual pigmentation SNPs showed strong associations with self-reported skin color, with a notable overlap, but not total congruence, with the SNPs associated with the M-index.

Conclusion

Our findings demonstrate that combinations of individual SNPs beyond those present in HIrisPlex-S, significantly improve the prediction of constitutive pigmentation compared to standard FDP tools and global ancestry alone. The role of income in "Pardo" classification, highlights the impact of social variables on phenotypic labeling. For forensic practice, these results underscore the necessity of population-specific approaches and a nuanced interpretation of DNA-led predictions in diverse social contexts. 

Financial support

FAPESP (16/03284-8), LIM40-HCFMUSP, LIM38-HCFMUSP.

Authors

• Carolina Bonilla (Departamento de Medicina Preventiva & Departamento de Dermatologia, Faculdade de Medicina, Universidade de São Paulo, Brazil)
• Alissa Ukei (Departamento de Medicina Preventiva, Faculdade de Medicina, Universidade de São Paulo, Brazil)
• Rafael Araujo (Departamento de Medicina Legal, Bioética, Medicina do Trabalho e Medicina Física e Reabilitação, Faculdade de Medicina, Universidade de São Paulo, Brazil)
• Cintia Fridman (Departamento de Medicina Legal, Bioética, Medicina do Trabalho e Medicina Física e Reabilitação, Faculdade de Medicina, Universidade de São Paulo, Brazil)