A Fast-Fourier-Transform-Based Dynamic Likelihood Ratio Framework for Controlling False Positives in DNA Database...

Area of Science:

• Forensic Science
• Computational Biology
• Genetics

Background:

• Traditional DNA databases use static locus-count thresholds, which can miss crucial leads from degraded samples or generate false matches with common alleles.
• Existing methods are computationally simple but lack the flexibility to handle the complexities of real-world forensic data and large-scale database comparisons.

Purpose of the Study:

• To introduce an automated framework for dynamic likelihood ratio (LR) thresholding in operational DNA databases.
• To overcome the limitations of static thresholds by implementing a more adaptive and accurate approach to match admissibility.

Main Methods:

• Employed a Fast Fourier Transform (FFT) algorithm to calculate the Probability Mass Function (PMF) for shared loci in real-time.
• Integrated the Balding-Nichols model to account for population substructure.
• Defined admissibility based on a user-defined maximum acceptable false positive rate at a specified confidence level, adapting the LR threshold to database size.

Main Results:

• The dynamic framework allows precise prediction and adaptation of search criteria to manage administrative workload.
• Massive-scale simulations across five population groups validated the approach, showing static thresholds collapse under large-scale comparisons.
• Receiver Operating Characteristic (ROC) and Poisson analyses demonstrated that static rules lead to unmanageable false positive risks when databases scale.

Conclusions:

• The dynamic framework offers a mathematically rigorous and scalable solution by linking decision thresholds to database size and genetic rarity.
• It successfully identifies rare, low-locus matches often discarded by static rules.
• Provides a method to maintain a predefined expected false positive rate, enhancing forensic DNA database efficiency and accuracy.