Paternity testing and other inference about relationships from DNA mixtures.

Area of Science:

• Forensic Science
• Genetics
• Statistical Modeling

Background:

• Interpreting DNA mixtures presents challenges due to sample quality and multiple contributors.
• Likelihood ratios are crucial for evaluating evidence in relationship testing.
• Existing statistical models for DNA mixtures can be computationally intensive.

Purpose of the Study:

• To develop and present methods for inferring relationships between DNA mixture contributors and known individuals.
• To enhance the accuracy of likelihood ratio calculations in forensic casework.
• To demonstrate the utility of Bayesian networks in DNA mixture interpretation.

Main Methods:

• Utilized a Bayesian network (BN) as a computational device for a statistical DNA mixture model.
• Applied the methods to real casework examples from criminal and paternity investigations.
• Incorporated additional genotype information to strengthen inferential power.

Main Results:

• The developed methods effectively evaluate evidence as likelihood ratios for specified relationships versus no relationship.
• Analysis of real casework demonstrated the practical application of the DNA mixture model.
• Taking full account of DNA mixture uncertainty yielded likelihood ratios comparable to single-source profiles.

Conclusions:

• The Bayesian network approach provides robust methods for relationship inference from DNA mixtures.
• The methods are flexible and extendable to various genotyping kits, databases, and hypotheses.
• Freely available R code facilitates the application of these advanced DNA analysis techniques.