BR-FDP-EYE

Forensic DNA phenotyping seeks to predict an individual's physical appearance from a DNA sample. Eye color, a highly heritable genetic trait, is a prominent and easily distinguishable characteristic useful for human identification. Genotype-phenotype association studies have identified single nucleotide polymorphisms (SNPs) within genes that directly or indirectly influence pigment synthesis.

The BR-FDP-EYE eye color prediction tool employs machine learning genetic analysis to predict eye color based on Southern Brazilian population DNA. By analyzing SNPs, the tool estimates the probability of different eye colors. Four machine learning models interpret genetic data to provide reliable predictions. This tool offers a user-friendly platform for forensic DNA phenotyping.

The BR-FDP-EYE prediction system analyzes 5 to 7 key SNPs associated with eye color. Upon sample data submission, the tool executes the selected analysis to determine the most probable eye color.

Prediction Models:

To ensure accurate and reliable eye color predictions, four distinct machine learning models were developed:

• BR-FDP-EYE-5: a 5-class Multilayer Perceptron (MLP) classifier for predicting all eye color categories: Blue, Green, Hazel, Light Brown, and Dark Brown.
• BR-FDP-EYE-3-I: a 3-class Support Vector Machine (SVM) classifier for distinguishing between Blue, Intermediate (Green, Hazel, Light Brown), and Dark Brown.
• BR-FDP-EYE-3-II: a 3-class K-nearest Neighbor (KNN) classifier for classifying between Light (Blue or Green), Hazel, and Brown (Light or Dark).
• BR-FDP-EYE-3-III: a 3-class KNN classifier for classifying between Blue, Intermediate (Green or Hazel), and Brown (Light or Dark).

These models were trained with varying class grouping strategies, offering users versatile prediction options tailored to their specific requirements.

If you use BR-FDP-EYE in a scientific publication, we would appreciate citations to the following paper:

BibTeX

@article{brfdpeye,
    author = {},
    title = {},
    journal = {},
    volume = {},
    number = {},
    pages = {},
    year = {},
    doi = {}
}