Mutation or uniparental disomy? Evaluating abnormal inheritance patterns using Object-Oriented Bayesian Networks.

Area of Science:

• Genetics
• Forensic Science
• Biostatistics

Background:

• Paternity testing traditionally assumes incompatibilities are paternal or due to mutations.
• Uniparental disomy (UPD) is increasingly recognized as a factor in genetic inheritance.
• Existing models may not adequately address complex inheritance patterns like UPD.

Purpose of the Study:

• To develop a statistical framework for paternity testing that incorporates both mutation events and UPD.
• To create an Object-Oriented Bayesian Network (OOBN) model for evaluating paternity trio results.
• To statistically assess likelihood ratios (LR) for paternity hypotheses, considering atypical inheritance.

Main Methods:

• Designed an OOBN model using GeNIe Academic software.
• Developed a computational tool in Python utilizing the pmgpy package.
• Simulated three million paternity trios to test the model's ability to identify and distinguish mutations and UPD.

Main Results:

• The OOBN model successfully calculates LR values, accurately accounting for mutation and UPD events.
• The framework provides unbiased statistical results, not assuming paternal origin of incompatibilities.
• The model can estimate the probabilities of mutation and UPD based on genetic data and population frequencies.

Conclusions:

• The proposed OOBN framework offers a robust method for paternity testing, especially when discrepancies suggest atypical inheritance.
• It improves the accuracy of paternity probability estimation by including UPD and mutation events.
• The computational tool enables scalable processing of large datasets for genetic parentage analysis.