Probabilistic Searching of Databases Compared to Binary CODIS Eligibility: An Improved Methodology for Maximizing Investigative Leads Using STR DNA Profiles

  • Session

Probabilistic Searching of Databases Compared to Binary CODIS Eligibility: An Improved Methodology for Maximizing Investigative Leads Using STR DNA Profiles

  • October 27, 2026 //
  • 2:01 pm -
  • 2:21 pm //
  • Rhode Island Convention Center

Probabilistic methods for DNA mixture interpretation have improved the ability to deconvolve complex mixtures for subsequent database searching. Current approaches typically rely on summarizing obligate alleles or genotypes using thresholds (e.g., ≥99% genotype weights), which can limit the information used in downstream comparisons.

Here, we evaluate a fully probabilistic search strategy to improve database searching performance. DBLR™ (PHF Science, New Zealand) is software that assigns likelihood ratios (LRs) to profiles within a database of STRmix™ deconvolutions, enabling direct comparison using all available information from a questioned mixture. In addition, DBLR™ can support databases larger than current CODIS implementations.

A total of 155 DNA mixtures, derived from 81 unrelated individuals across eight laboratories, were examined. These 2–4 person mixtures varied in mixture proportion, DNA input, PCR cycles, and STR kits. The mixtures were previously deconvolved in STRmix™ (PHF Science, New Zealand) using each laboratory’s validated parameters, resulting in 765 deconvolutions and 1,220 contributor “slots” available for comparison.

Deconvolution results were first evaluated using a CODIS Moderate Match Estimate (MME) proxy to determine eligibility for upload (threshold: 1 in 10,000,000). The same deconvolutions were then searched using DBLR™ against a database of 25,000,081 profiles (25 million non-donors and 81 ground truth known donors) with an LR threshold of 10,000,000.

Only 37.7% of comparisons met the CODIS MME threshold for database upload. In contrast, DBLR™ identified 83.4% of true donors above the LR threshold while producing only four non-donor hits.

These results demonstrate that probabilistic LR-based database searching can substantially increase the number of detectable true donors compared to current binary eligibility criteria, while maintaining a low rate of adventitious matches. This approach offers a promising alternative for maximizing investigative leads from complex DNA mixtures, particularly for profiles that fail to meet existing CODIS requirements.

Probabilistic methods for DNA mixture interpretation have improved the ability to deconvolve complex mixtures for subsequent database searching. Current approaches typically rely on summarizing obligate alleles or genotypes using thresholds (e.g., ≥99% genotype weights), which can limit the information used in downstream comparisons.

Here, we evaluate a fully probabilistic search strategy to improve database searching performance. DBLR™ (PHF Science, New Zealand) is software that assigns likelihood ratios (LRs) to profiles within a database of STRmix™ deconvolutions, enabling direct comparison using all available information from a questioned mixture. In addition, DBLR™ can support databases larger than current CODIS implementations.

A total of 155 DNA mixtures, derived from 81 unrelated individuals across eight laboratories, were examined. These 2–4 person mixtures varied in mixture proportion, DNA input, PCR cycles, and STR kits. The mixtures were previously deconvolved in STRmix™ (PHF Science, New Zealand) using each laboratory’s validated parameters, resulting in 765 deconvolutions and 1,220 contributor “slots” available for comparison.

Deconvolution results were first evaluated using a CODIS Moderate Match Estimate (MME) proxy to determine eligibility for upload (threshold: 1 in 10,000,000). The same deconvolutions were then searched using DBLR™ against a database of 25,000,081 profiles (25 million non-donors and 81 ground truth known donors) with an LR threshold of 10,000,000.

Only 37.7% of comparisons met the CODIS MME threshold for database upload. In contrast, DBLR™ identified 83.4% of true donors above the LR threshold while producing only four non-donor hits.

These results demonstrate that probabilistic LR-based database searching can substantially increase the number of detectable true donors compared to current binary eligibility criteria, while maintaining a low rate of adventitious matches. This approach offers a promising alternative for maximizing investigative leads from complex DNA mixtures, particularly for profiles that fail to meet existing CODIS requirements.

  • Early Registration
  • $250.00
  • Standard Registration (after July 15)
  • $295.00
  • Student Registration$195.00
  • $195.00

Workshop currently at capacity. A waitlist is available to join on our registration page.

Brought to you by

Worldwide Association of Women Forensic Experts

Michael Coble

Professor and Executive Director, Center for Human Identification (CHI), University of North Texas Health Science Center

Michael Coble, PhD, is a Professor and the Executive Director of the Center for Human Identification at the University of North Texas Health Fort Worth. Dr. Coble received his PhD in Genetics from The George Washington University. He has over 100 peer-reviewed publications in forensic DNA analysis and interpretation and is recognized among the top 2% of highly cited researchers worldwide.

Speaker Image

Submit Question to a speaker