Regression models for DNA-mixtures.

Area of Science:

• Forensic Science
• Genetics
• Statistical Modeling

Background:

• Conventional forensic DNA analysis relies on Short Tandem Repeat (STR) markers.
• Increasing the number of markers, such as Single Nucleotide Polymorphism (SNP) markers, can enhance the power of DNA mixture interpretation.
• New statistical approaches are required to effectively utilize SNP data for complex DNA mixtures.

Purpose of the Study:

• To develop and validate a novel regression framework for the statistical interpretation of DNA mixture evidence.
• To replace traditional hypotheses (HD and HP) with parametric versions based on contributor fractions.
• To assess the model's performance in accurately estimating contributions and detecting low-level DNA in mixtures.

Main Methods:

• A parametric regression model was developed, directly utilizing DNA peak heights.
• The model estimates the contributed fraction of each person in a mixture, treating it as a regression parameter.
• The framework was tested on 25 controlled experiments with 2-5 contributors and validated through simulations.

Main Results:

• The regression model accurately estimated contributor fractions from 0.01 to 0.99.
• No false positives were observed, meaning non-contributors were not identified as contributors.
• Some false negatives occurred for contributions at or below 0.1, but simulations showed reliable detection of 1% contributions with sufficient SNP markers.

Conclusions:

• The proposed regression framework offers a powerful statistical tool for DNA mixture interpretation using SNP markers.
• The method effectively handles issues like drop-in, drop-out, and an unknown number of contributors.
• This approach has the potential to analyze cases previously beyond the capabilities of conventional forensic methods, especially with a larger number of SNP markers.