417 F.Supp.3d 857
W.D. Mich.2019Background
- Defendant Daniel Gissantaner charged with being a felon in possession of a firearm; prosecution’s case rests mainly on touch‑DNA recovered from a gun found in a locked cedar chest.
- MSP STRmix™ analysis reported Gissantaner as a minor (≈7%) contributor and produced a likelihood ratio (~49 million) favoring inclusion over three unknown contributors.
- Evidence and chain‑of‑custody handling had irregularities; the DNA sample was a low‑template, multi‑person touch mixture (~0.7 ng total, ~49 pg from minor contributor).
- Extensive Daubert proceedings: government and defense experts testified; Court appointed independent experts (Drs. Coble and Krane) to evaluate STRmix and MSP validation.
- Key factual and technical disputes: adequacy of MSP’s internal validation for low‑template 3+ person mixtures, operator/laboratory input variability, limited independent peer review and industry standards, and potential error/interpretation risks.
- Outcome: District Court excluded the STRmix‑derived likelihood ratio under Daubert, concluding the evidence was not shown reliable in these circumstances.
Issues
| Issue | Plaintiff's Argument | Defendant's Argument | Held |
|---|---|---|---|
| Admissibility of STRmix results under Daubert/Rule 702 | Reliability, not perfection; STRmix widely used; court‑appointed expert Coble endorses application here | STRmix output depends on subjective/operator inputs, software and lab practices not adequately validated for the sample | Excluded: government failed to prove reliability for this case’s sample |
| Adequacy of MSP testing/validation for STRmix on this sample | MSP performed internal validation; other labs use STRmix; later MSP case studies encompassed similar conditions | MSP validation lacked clear limits for low‑template, minor‑contributor mixtures and omitted key validation data; external review limited | Excluded: validation insufficient for the low‑template, 3+ contributor, ~7% minor contributor sample |
| Peer review and general acceptance | Numerous peer‑reviewed articles and many labs employing STRmix show scientific engagement and community adoption | Many publications authored by developers; independent, multidisciplinary (incl. software engineering) peer review limited | Mixed: general acceptance for mainstream/higher‑quality mixtures, but not persuasive for low‑template complex mixtures at issue |
| Error rates, standards, and software development practices | Error risk mitigated by analyst review/diagnostics; SWGDAM guidance exists and ASB/OSAC standards in progress | No binding standards or comprehensive V&V per software‑engineering norms; unknown false‑positive/negative behavior at sample’s margin; risk of misinterpretation (e.g., prosecutor's fallacy) | Weighed against admissibility: lack of standards and unclear error characterization undermined reliability |
Key Cases Cited
- Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (establishes reliability gatekeeping for expert testimony under Rule 702)
- Kumho Tire Co. v. Carmichael, 526 U.S. 137 (gatekeeping applies to all expert technical/specialized testimony)
- People v. Collins, 15 N.Y.S.3d 564 (N.Y. Sup. Ct. 2015) (example of a court rejecting probabilistic genotyping admissibility)