Forensic DNA Phenotyping (FDP) tools, like the HIrisPlex-S system, infer externally visible characteristics from biological traces. However, in admixed populations like the Brazilian , skin pigmentation categorization involves complex biological and sociocultural variables, demanding rigorous validation of predictive methodologies against true phenotypes. Subjective classifications, such as self-report, are heavily influenced by cultural constructs and skin-color bias, frequently diverging from actual biological pigmentation. Therefore, this study investigates the correlation between self-reported skin color, objectively quantified melanin density, and HIrisPlex-S genetic predictions. A cross-sectional study was conducted with 611 unrelated Brazilian individuals. Constitutive melanin density (M-Index) was quantified using reflectance spectrophotometry, while subjective phenotypic data were obtained through skin color self-report. Genotyping of HIrisPlex-S pigmentation SNPs generated predictive probabilities for each color category. The association between variables was analyzed using Analysis of Variance (ANOVA) to assess melanin density across self-declared categories, and Pearson's test to correlate melanin with genetic predictions. The sample (N = 611) presented a mean constitutive melanin (M-Index) of 4.65 ± 1.02. Self-reported skin color comprised 366 "Brancos" (White), 158 "Pardos" (Brown), 50 "Pretos" (Black), and 37 "Amarelos" (Asian). Self-reported "Pardos" exhibited a higher mean M-Index (5.14 ± 1.11) compared to "Pretos" (4.79 ± 0.77). When correlating HIrisPlex-S probabilities with self-reported skin color, predictions for Very Pale and Pale skin showed negative correlations with the "Pardo" and "Preto" groups (r = -0.304 and r = -0.465, respectively; p < 0.001). Conversely, Dark and Dark-to-Black predictions showed positive correlations with these categories (r = 0.365 and r = 0.499, respectively; p < 0.001). Actual melanin density correlated with genetic predictions (Pale: r = -0.321, p < 0.001; Dark: r = 0.204, p < 0.001), but lacked significant correlation for the Intermediate skin prediction (r = 0.006, p = 0.871). The differences between objective melanin measurements and subjective self-report show how social factors, such as colorism, influence skin color categorization in Brazil. Although the HIrisPlex-S system works well for extreme phenotypes, it fails to predict intermediate skin colors accurately. This indicates that predictive models developed for more homogeneous populations lose their accuracy when applied to highly diverse groups. Consequently, relying solely on imported forensic panels or self-reported information could mislead forensic investigations. To improve the reliability of Forensic DNA Phenotyping, it is necessary to develop new predictive models specifically adapted to the Brazilian population. Financial Support: FAPESP (16/03284-8), LIM40-HCFMUSP

Authors

• Cintia Fridman (Faculdade de Medicina, Universidade de São Paulo, São Paulo, SP, BR, Brazil)
• Rafael Araujo (Faculdade de Medicina, Universidade de São Paulo, São Paulo, SP, BR, Brazil)
• Alissa Ukei (Faculdade de Medicina, Universidade de São Paulo, São Paulo, SP, BR, Brazil)
• Carolina Bonilla (Faculdade de Medicina, Universidade de São Paulo, São Paulo, SP, BR, Brazil)