STRmix is a software for the interpretation of complex DNA profiles. The three developers are Dr Duncan Taylor (Forensic Science South Australia), Dr Jo-Anne Bright and myself from ESR in New Zealand. We do not receive any benefit direct or indirect from sales of STRmix. We are all civil servants on the pay of our respective states.
The proceeds fro STRmix are shared between our US based partners, our development partners, and ESR. No profits are paid to any persons in new Zealand or Australia.
A brief history of STRmix
STRmixTM was developed in response to a lab closure of the Victoria Police Forensic Science Laboratory (VPFSL) in Melbourne, Australia in 2010. VPFSL had been using a software DNAmix which was created as a result of mixture work for the OJ Simpson case undertaken by Bruce Weir and John Buckleton. The software had been enhanced and upgraded by G Beecham. DNAmix should not be applied to loci where drop-out is possible. If loci that may exhibit drop-out are ignored there is still a minor risk of overstating the value of the evidence as the omitted loci may favor the proposition that the POI is not a donor. The situation which triggered the closure of the lab was the reissue of a statement where the original likelihood ratio had been given as 550 billion and the reissued statement gave the LR as 3. The analyst in this case had signaled to VPFSL management that this change was likely to cause concern. It triggered the closure of the lab.
Consequential on this closure an Australian and New Zealand (Australasian) standardization project was launched. At the time of this launch, labs in Australasia used a mix of manual methods for the assignment of likelihood ratios, freeware, in house EXCEL sheets, and the Tasmanian lab still used CPI. TrueAllele was being investigated as a potential tool and the New South Wales lab, then called DAL but now called FASS, had purchased TrueAllele. The standardization project had various components but of relevance to this text it was decided to build an in house software to serve as the “workhorse” for simple cases whilst TrueAllele would be used for more complex cases.
Perlin had initially been experimenting with a non-MC approach termed Linear Mixture Analysis,  subsequently patented but never advanced into a tool. As part of a tool, termed TrueAllele, developed for the UK Forensic Science Service DNA Databank Unit, Perlin had developed a baseline and noise management approach. Sometime prior to about 2009  Perlin developed his MCMC tool termed TrueAllele casework. TrueAllele casework inherits the baseline and noise management aspect.
Duncan Taylor from the Forensic Science South Australia laboratory (FSSA) and John Buckleton from ESR, New Zealand were initially tasked with producing a tool based on the drop-model that could deal with two person mixtures with no consideration of stutter. This occurred in May 2011. Initial efforts focused on analytical transformations of peak heights to heterozygote balance and the associated Jacobians. In the first week of development Ross Vining, then director of FSSA, suggested that such analytical transformations had been superseded by big computing. Vining was soon to die prematurely piloting his own small plane. However he had sparked a new direction for the project. At this stage Taylor and Buckleton turned to MCMC. David Balding, then resident in England but now Professor at the University of Melbourne, provided invaluable advice in the application of the Metropolis-Hastings algorithm which was to become the centerpiece of the mathematics in STRmix.
Taylor had created the core of a software that could do four person mixtures with stutter. Even the limit to four person mixtures was artificial, imposed by Buckleton. At this stage the third developer Jo-Anne Bright was added to the team.
Since this time improvements have been significant but incremental.
The first use in casework was August 2012 at Adelaide and New Zealand.
Initially, STRmixTM was not intended for international release. In late 2012 Dr Timothy Kalafut of the US Army Criminal Investigation Laboratory (USACIL) began a correspondence that led to a visit to New Zealand by 10 members of USACIL and NicheVision in April 2013 to learn about STRmix. This exposure led Steven Myers of California Department of Justice to contact us and he visited New Zealand to be trained and use the software in September 2013. Also attending at this time was Dr Ian Evett from the UK and subsequently Dr Mike Coble from the National Institute of Standards and Technology.
After much introspection STRmixTM was released for international use in January 2014.
It is currently in use in 30 labs in the US, all 8 State and territory labs in Australasia, and 5 labs elsewhere. See here for a full list of labs using STRmix in caseworkLabs live
A list of the peer reviewed publications for STRmix appears here
Peer reviewed publications for STRmix
A summary of the identified miscodes in STRmix appears here A summary of the seven identified miscodes in STRmix
People v Hillary People v Hillary II
A summary of the commercial releases all of which underwent full developmental validation appears here A summary of the commercial versions all of which had a full developmental validation
A non-exhaustive list of US cases:
NY v Bullard-Daniels , Niagara County strmix-niagara-county-ny-decision-_201603101415
State of Texas vs. Michael Shane Clack. Daubert/admissibility hearing April/20/16,
State of Texas vs. Roy Edward Smith April/20/16 Smith County near the Dallas area.
Appeal Roy Edward Smith v Texas 16-139-CR Smithv StateApr 28’17
Henry Watkins SKINNER, Appellant v. The STATE of Texas 8th June 2016 Court of Criminal Appeals of Texas 2016 WL 3351308 Skinner v State Tex_ Court of Criminal Appeals 2016 – Google Scholar
Roy Edward SMITH, Appellant v. The STATE of Texas, Appellee 28th April 2016 Court of Appeals of Texas, Tyler 2017 WL 1534048
People v. Irby (heard in Michigan). June 21st 2016
People v Sayers (Mi) 9th November 2016,
Daubert People v Herbert Alford (Mi) 16th November 2016,herbert-alford-opn-and-order-re-strmix-m-in-limine-11-28-16
Florida vs. Marc Regisme Daubert hearing denied 22nd November 2016, state_v-_regisme-daubert-denial
United States of America,v. Kenneth Pettway, Jr. and Demetrius Black, Defendants. 21st October 2016 Western District New York 2016 WL 6134493
People v Hillary (NY) hillary-08-26-16-decision-and-order-dna-analysis-admissibility, 07-25-16-fye-hearing-transcript
US v. Justin VAZQUEZ December 30th 2016. UNITED STATES OF AMERICA v Justin VAZQUEZ Defendant
Michigan v Larry David Smith 3rd May 2017 Smith Daubert Ruling
Florida v Dwayne Cummings 12th May 2017 State of Florida v Dwayne Cummings Daubert Denied
Daubert Michigan Genesee County People v Marlon Burns 15th June 2017 STRmix admitted Daubert Order Burns
Texas v Nelson 17th August 2017 Nelson v State
Daubert Wyoming v Bradley Ross Fairbourn 25th August 2017 admitted FILELaunch718756_14737
Minnesota v Johnny Earl Edwards 31st October, 2017 admitted Atty-Admin6-SharpMX753N-DoNotReply@co.anoka.mn.us_20171115_161036
California v Alvin Larry Davis Kelly Hearing Stockton, Ca 18-19th December 2017 Admitted Buckleton_Davis_Testimony
US v Melvin Russell, US District Court for the district of New Mexico. 31st October, 2017. Daubert hearing, evidence admitted. Melvin Russell
R v Pfening (Australia) r-v-pfennig-judgement-11-nov-2016
R v Sandra Weir (Scotland) for the murder of Mary Logie High Court in Edinburgh 5th January 2017. Mr report was agreed with the defence who were Beltrami solicitors. The Scottish Police Agency scientist was asked me in court why they had sent work to ESR, how we sent the data, how the software works and if it is accepted in the scientific community
Josephus Marinus Johannes de Graef 21st November 2016
Due to the difficult LCN-profiles (34 cycles Identifiler, three replicates) this was a very special case. It was the first case in the Netherlands where the DNA-evidence was discussed in such detail. The probabilistic software and the outcome of three programs (LRmix, TrueAllele, STRmix) was discussed.
The judges found the DNA-evidence of STRmix and LRmix/ Mixcall (in-house developed version of LRmix) reliable. In the link below, under chapter “7.3.1 Weergave van de likelihood ratio’s.” you can find a table with results of Netherlands Forensic Institute and STRmix. They are similar. TrueAllele gave different result.
TrueAllele was run on only one replicate although be capable of being run on all three. The effect of replication is very considerable especially at low-template. Hence it would be unfair to score the TrueAllele result against the other two.
July 2017 R v Naweed Ali, Khobaib Hussain, and Mohibur Rahman at the Central Criminal Court (the ‘Old Bailey’) STRmix admissibility challenged. Cellmark provided validation documents and details of previous court challenges, including supporting information provided by ESR. The evidence was admitted. http://www.bbc.co.uk/news/uk-40802787
Nelson v Texas This is not a STRmix case but STRmix features https://scholar.google.co.nz/scholar_case?case=15999635421785956074&q=strmix+dna&hl=en&as_sdt=2006
The views expressed in this site are my own and do not necessarily represent those of my organisation.