United States v. Williams
382 F. Supp. 3d 928
N.D. Cal.2019Background
- Defendant Reginald Elmore is charged with an August 14, 2008 double murder; trial set for May 2019. The motion concerns DNA mixture evidence from the right rear door handle of the rental car.
- SERI first tested the car swabs in 2012 (Harmor, Identifiler Plus, 15 loci) and concluded the door-handle swab contained DNA from at least five contributors.
- In 2018 the government requested retesting; Phillip Hopper (SERI) reanalyzed using GlobalFiler (21 loci) and the probabilistic genotyping program Bullet, concluding the mixture was four contributors and Bullet produced a likelihood ratio strongly supporting Elmore as a contributor.
- SERI validated Bullet only for two-, three-, and four-person mixtures; it did not validate Bullet for five- or six-person mixtures.
- Scientific literature and SERI’s own validation showed high rates of underestimating five- and six-person mixtures as four-person mixtures; SERI’s GlobalFiler validation misidentified all known five-person mixes as four-person.
- The sample tested by Hopper had limiting features: low quantity of DNA (0.65 ng, below recommended 1 ng), signs of degradation since 2012, and two loci with seven alleles (one with a below-threshold possible eighth allele), increasing risk that contributors were underestimated.
Issues
| Issue | Plaintiff's Argument | Defendant's Argument | Held |
|---|---|---|---|
| Whether Hopper's Bullet-based opinion is admissible under Rule 702 because Bullet was not validated for mixtures with >4 contributors | Elmore: Bullet was not validated for five+ contributor mixtures and SERI has a high error rate underestimating such mixtures; Hopper did not reliably establish the mixture had only four contributors, so his Bullet analysis is unreliable | Government: Underestimation risk does not prove this particular mixture was mischaracterized; Hopper’s four-contributor conclusion and Bullet result are admissible and subject to cross-examination | Court: Granted motion to exclude. Bullet not validated for this mixture because Hopper did not reliably establish only four contributors; foundational input (number of contributors) is unreliable, so the analysis is inadmissible |
| Whether the government could instead present Bullet results under alternative contributor assumptions (e.g., five contributors) | Elmore: Bullet is not validated for five-person analyses and those outputs are unreliable and inadmissible | Government: Bullet can be run under a five-person assumption for comparison; result did not change dramatically | Court: Rejected reliance on the five-person Bullet output because SERI did not validate Bullet for five-person analyses and Hopper would not rely on that statistic |
Key Cases Cited
- Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579 (admissibility turns on relevance and reliability of expert testimony)
- Cooper v. Brown, 510 F.3d 870 (expert testimony must assist the trier of fact)
- Primiano v. Cook, 598 F.3d 558 (Rule 702 reliability and relevance framework; flexible inquiry)
- i4i Ltd. P'ship v. Microsoft Corp., 598 F.3d 831 (when methodology is sound, disputes about accuracy go to weight)
- Sandoval-Mendoza, 472 F.3d 645 (reliability may focus on personal knowledge or experience)
- Lust v. Merrell Dow Pharm., 89 F.3d 594 (proponent bears burden to show admissibility by preponderance)