Shotgun DNA sequencing evidence: Sample-specific and unknown genotyping error probabilities.

Area of Science:

• Forensic Genetics
• Computational Biology
• Statistical Modeling

Background:

• Degraded DNA in forensic trace samples (e.g., telogen hairs) often prevents short tandem repeat (STR) profiling.
• Shotgun DNA sequencing offers an alternative, yielding valuable single nucleotide polymorphism (SNP) data from low-quality samples.
• Accurate statistical models are essential for interpreting shotgun sequencing data in forensics, including accounting for sequencing errors.

Purpose of the Study:

• To extend the wgsLR statistical model for forensic DNA analysis.
• To incorporate asymmetric genotyping error probabilities between trace and reference samples.
• To address unknown genotyping error probabilities using profile likelihood maximization and prior distributions.

Main Methods:

• Extension of the existing wgsLR model to accommodate differing error rates.
• Implementation of methods to estimate unknown genotyping error probabilities.
• Investigation of model robustness against overdispersion and prior distribution specifications.
• Development of an R package (wgsLR) for the extended model.

Main Results:

• The extended wgsLR model successfully handles asymmetric genotyping error probabilities.
• Unknown genotyping error probabilities can be reliably estimated with sufficient independent markers.
• The model demonstrates robustness to different prior distribution specifications.
• The model is robust against overdispersion, enhancing its reliability in forensic applications.

Conclusions:

• The enhanced wgsLR model provides a more accurate and robust framework for evaluating forensic DNA evidence from shotgun sequencing data.
• The ability to handle unknown and asymmetric error rates improves the interpretation of low-quality trace samples.
• The R package wgsLR facilitates the practical application of these advanced statistical methods in forensic genetics.