Back to Presentations

From Match Strength to Match Accuracy: Answering the Ultimate DNA Question

M. W. Perlin, W. Allan, J. Bracamontes, M. Legler, "From Match Strength to Match Accuracy: Answering the Ultimate DNA Question", Promega's Thirty Seventh International Symposium on Human Identification, Providence, RI, 28-Oct-2026.

Poster

Download Poster

Abstract

Forensic DNA interpretation has traditionally focused on match strength. After inferring an unknown genotype from evidence, we compare that genotype with a reference profile and report a likelihood ratio (LR) that measures the strength of the comparison. This approach has become a cornerstone of modern forensic science and is now widely used for complex DNA evidence.

Yet an important question remains largely unanswered. Before comparing an inferred genotype with a suspect, victim, database profile, or relative, how reliable is that genotype itself? How well can it distinguish a true contributor from a noncontributor? This presentation introduces a new framework for answering that ultimate question.

Previous research had developed match strength distributions derived from uncertain genotypes (Perlin, 2018). These paired distributions describe how an inferred genotype behaves across all possible match and nonmatch comparisons. Rather than focusing on a single LR value, they characterize the overall identification performance of the genotype.

The new framework applies receiver operating characteristic (ROC) analysis in a new way to measure genotype discrimination accuracy (Perlin et al, 2026). This novel high-resolution ROC analysis combines the genotype’s match and nonmatch LR distributions into a simple and intuitive probability measure of discrimination accuracy.

The area under the ROC curve (AUC) can be interpreted as the probability that a true contributor will receive stronger support than a noncontributor. In practical terms, the AUC measures how accurately an inferred genotype can distinguish the people who contributed DNA from those who did not.

This classification perspective provides a new way to evaluate forensic DNA evidence. Instead of asking only how strongly a genotype matches a particular individual, we can ask how accurately that genotype can make match decisions in general. The answer is a single probability number that quantifies the identification capability of the evidence itself.

The approach has practical applications throughout forensic DNA interpretation. It can be used to assess genotype quality, compare analytical methods, evaluate the impact of thresholds, measure information loss, and understand why some evidence profiles are more informative than others. It also provides a quantitative framework for validation studies and quality assessment by directly measuring the discrimination performance of inferred genotypes.

Beyond these applications, the framework suggests a broader shift in how forensic DNA evidence can be understood. Traditionally, the inferred genotype has been viewed primarily as an intermediate result used to calculate LR values. In the new framework, the genotype becomes the central scientific result. From that genotype, one can derive match-strength distributions, estimate error rates, measure discrimination accuracy, and evaluate identification performance.

The likelihood ratio evaluates a person against a genotype. ROC discrimination accuracy evaluates the genotype itself. These measures are complementary, but they answer different questions. The former quantifies the strength of a particular comparison. The latter quantifies the reliability of the underlying genotyping data and inference.

For decades, forensic DNA has focused on measuring match strength. This new work introduces a practical way to measure match accuracy. By quantifying how well an inferred genotype can distinguish contributors from noncontributors, discrimination probability provides a direct answer to a fundamental forensic question: how reliable is the DNA evidence as a tool for identification?

References

1. Perlin MW. Efficient construction of match strength distributions for uncertain multi-locus genotypes. Heliyon. 2018;4(10):e00824.
2. Perlin MW, Allan WP, Bracamontes JM, Legler MM. The reliability and reporting of DNA match strength for uncertain genotype evidence. Heliyon. 2026;12:e44753.

Links

• Promega's Thirty Seventh International Symposium on Human Identification - Website