STRmix / Mixture Interpretation Review

Independent review of STRmix likelihood ratios, assumptions, and mixture interpretation

STRmix is probabilistic genotyping software used to interpret DNA profiles – particularly mixed DNA profiles – and produce likelihood ratios (LRs).

It applies statistical modelling to account for common DNA profile behaviours such as mixture composition, peak height variation, drop-in, drop-out, degradation and conditioning assumptions, using large numbers of computational simulations.

Importantly, the likelihood ratio produced by STRmix is not a likelihood that a person touched the item, nor that their DNA was deposited during the alleged events. It is a source-level statistic, and its meaning depends on the propositions selected.

STRmix has been extensively validated for forensic use. However, like all interpretive frameworks, it involves assumptions and limitations that must be understood and tested when evaluating the strength of a reported result.

 

 

STRmix is probabilistic genotyping software used to interpret DNA profiles – particularly mixed DNA profiles – and produce likelihood ratios (LRs).

It applies statistical modelling to account for common DNA profile behaviours such as mixture composition, peak height variation, drop-in, drop-out, degradation and conditioning assumptions, using large numbers of computational simulations.

Importantly, the likelihood ratio produced by STRmix is not a likelihood that a person touched the item, nor that their DNA was deposited during the alleged events. It is a source-level statistic, and its meaning depends on the propositions selected.

STRmix has been extensively validated for forensic use. However, like all interpretive frameworks, it involves assumptions and limitations that must be understood and tested when evaluating the strength of a reported result.

 

 

We assist criminal lawyers by:

  • Reviewing STRmix mixture interpretation and reported likelihood ratios (LRs)
  • Evaluating the strength and limitations of the statistical result
  • Identifying key modelling assumptions
  • Preparing clear advice to support case strategy
  • Delivering expert reports where required

Common issues seen in STRmix reporting:

  • Contributor number  assumptions
  • Low-level DNA and complex peak behaviour (drop-in, drop-out, degradation)
  • Conditioning assumptions 
    Source-level likelihood ratios being overstated
  • Limited transparency around modelling decisions, thresholds, and sensitivity

Contact us when:

  • The LR is low (e.g. below 10,000)
  • The DNA result is pivotal to the prosecution or defence case
  • The profile is low-level, degraded, or a complex mixture
  • Relatives (or close genetic relationships) are a realistic consideration
  • The LR appears unusually strong given the alleged circumstances

Depending on your needs and timeframes, engagement may include:

  • Consider written advice for solicitor/counsel
  • Conference notes addressing key interpretive issues
  • Cross-examination support (targeted questions and concessions)
  • Expert witness report (where required)
  • Ongoing support through committal, pre-trial, or trial preparation

If you are unsure whether the forensic evidence warrants expert involvement, I can provide an initial view and a quote for the appropriate level of review – from targeted advice through to full reporting and trial support.

What we do

We assist criminal lawyers by:

  • Reviewing STRmix mixture interpretation and reported likelihood ratios (LRs)
  • Evaluating the strength and limitations of the statistical result
  • Identifying key modelling assumptions
  • Preparing clear advice to support case strategy
  • Delivering expert reports where required
Common issues

Common issues seen in STRmix reporting:

  • Contributor number  assumptions
  • Low-level DNA and complex peak behaviour (drop-in, drop-out, degradation)
  • Conditioning assumptions 
    Source-level likelihood ratios being overstated
  • Limited transparency around modelling decisions, thresholds, and sensitivity
When to use us

Contact us when:

  • The LR is low (e.g. below 10,000)
  • The DNA result is pivotal to the prosecution or defence case
  • The profile is low-level, degraded, or a complex mixture
  • Relatives (or close genetic relationships) are a realistic consideration
  • The LR appears unusually strong given the alleged circumstances
What you get

Depending on your needs and timeframes, engagement may include:

  • Consider written advice for solicitor/counsel
  • Conference notes addressing key interpretive issues
  • Cross-examination support (targeted questions and concessions)
  • Expert witness report (where required)
  • Ongoing support through committal, pre-trial, or trial preparation
Quotation

If you are unsure whether the forensic evidence warrants expert involvement, I can provide an initial view and a quote for the appropriate level of review – from targeted advice through to full reporting and trial support.

GET A QUOTE

False inclusions in STRmix mixtures

In mixture interpretation, the key risk is often a false inclusion - a person being reported as included under a particular set of assumptions, despite uncertainty in mixture composition or low-level profile behaviour.

This is most likely in complex or low-template profiles where modelling decisions, conditioning assumptions and proposition wording materially influence the LR. Reviewing those assumptions is critical to ensuring the statistic is not overstated in court.

This risk can be further increased where close familial relationships are a realistic possibility, because relatives may share alleles that can make inclusion more likely in complex mixtures. In these circumstances, careful consideration of propositions, conditioning assumptions, and alternative contributors is critical.

Helen Roebuck DNA expert giving evidence

STRmix results require more than a number

Mixture interpretation is not mechanical.

Even with advanced software, results are shaped by:

  • the propositions selected
  • mixture complexity and contributor assumptions
  • conditioning information
  • low-level profile behaviour and uncertainty

My role is to assess whether the statistic has been produced and communicated in a way that is scientifically robust and legally meaningful and to ensure the evidence is not extended beyond what it can actually support.

Common issues seen in STRmix reporting

STRmix results can be presented in a way that appears definitive. In practice, careful review often identifies issues such as:

  • reliance on narrow or assumed propositions

  • the statistic being interpreted as “contact” or “involvement”

  • strong language that moves beyond source-level inference

  • mixture complexity not being communicated clearly

  • insufficient explanation of limitations, uncertainty, or modelling decisions

  • reporting that does not assist the court to understand what the LR does (and does not) mean

 

A strong LR may be technically correct – and still be misleading if used to support a conclusion it cannot logically answer.

STRmix results can be presented in a way that appears definitive. In practice, careful review often identifies issues such as:

  • reliance on narrow or assumed propositions

  • the statistic being interpreted as “contact” or “involvement”

  • strong language that moves beyond source-level inference

  • mixture complexity not being communicated clearly

  • insufficient explanation of limitations, uncertainty, or modelling decisions

  • reporting that does not assist the court to understand what the LR does (and does not) mean

 

A strong LR may be technically correct – and still be misleading if used to support a conclusion it cannot logically answer.

STRmix False Inclusion

A false inclusion provides inclusionary support of an individual, when that individual has not contributed to the DNA profile.
A STRmix false inclusion can be considered on factors including inputs, outputs, diagnostics and propositions.
Significant contributing factors to STRmix false inclusions are the subjective decisions that the STRmix analyst makes when conducting the interpretation.
An empirical study using STRmix simulated 300,000 mixed DNA profiles with false inclusions occurring on 2900 occasions.
Known contributing factors of false inclusion are DNA profile complexity, number of contributors, and allele sharing.

DNA Likelihood ratio

The DNA likelihood ratio (LR) is an internationally recognised numerical method to evaluate mixed DNA profiles.
Various methods can be used to formulate LR, including STRmix.
The likelihood ratio number could range up to 100 billion, and greater in certain jurisdiction.
The LR is a consideration of the likelihood of obtaining the observed DNA profile given two hypotheses, which should be the prosecution and defense propositions,
Valid DNA likelihood ratio calculation can turn on analyst inputs, including choice of propositions, allele frequency rate, theta correction, ethnicity considerations and relatives of the individual.

DNA Contamination

DNA contamination can affect evidence from crime scene, laboratory and throughout the entire chain of custody.
Such contamination can impact upon the STRmix assessment.
Increased testing sensitivity means the majority of DNA profiles contain levels of background DNA. Background DNA will be analysed during the STRmix assessment and is commonly defined as an unknown contributor.
Background DNA of unknown contributors can represent contamination.
A rigorous review of the STRmix outputs, along with the laboratory contamination report can assist in the identification of any contamination that may have contributed to the STRmix assessment.

STRmix Error Rate

STRmix error rate, known STRmix miscoding, and the variable nature of STRmix outputs, have been the subject of divisive scientific and legal debate through the evolution of STRmix and other PG (probabilistic genotyping) software development.
Multiple STRmix likelihood ratios will commonly be generated, and the analyst must exercise skill in the required subjective selection.
STRmix error reduction is reliant upon the appropriate formulation of propositions, contributor assigned and parameter settings, amongst other factors.
There are currently no published guidelines on the acceptable STRmix error rate.

DNA Ethnicity database

Valid DNA likelihood ratio calculation relies upon the appropriate ethnicity database selection.
In Australia, commonly selected datasets are Australian Aboriginal, Caucasian, Asian and Sudanese, though there are many others that can be utilised.
The numerical value (ie 100 billion) assigned to a DNA profile, requires an estimation as to how “common” or “rare” the profile is.
If the same DNA profile is assessed by STRmix using three different ethnicity databases’, it will produce three different likelihood ratios.
Certain laboratories have a tendency of applying a specific ethnicity, and others approach the matter on a case-by-case basis.

Number of Contributors

A primary limitation of STRmix is that the analyst estimates the number of contributors.
Incorrect estimation of the number of contributors can directly cause an invalid likelihood ratio.
Contributor selection is critically reliant upon the DNA analyst’s skill, experience, and ability to evaluate factors of allele sharing, allele dropout, and stutter.
To minimise risk of bias, best practice state that the analyst conducts contributor selection without consideration to the person of interest samples.
Interrogation of the STRmix primary and secondary diagnostic outputs can assist to determine if the likelihood ratio is valid.
False inclusion

STRmix False Inclusion

A false inclusion provides inclusionary support of an individual, when that individual has not contributed to the DNA profile.
A STRmix false inclusion can be considered on factors including inputs, outputs, diagnostics and propositions.
Significant contributing factors to STRmix false inclusions are the subjective decisions that the STRmix analyst makes when conducting the interpretation.
An empirical study using STRmix simulated 300,000 mixed DNA profiles with false inclusions occurring on 2900 occasions.
Known contributing factors of false inclusion are DNA profile complexity, number of contributors, and allele sharing.
Likelihood ratio

DNA Likelihood ratio

The DNA likelihood ratio (LR) is an internationally recognised numerical method to evaluate mixed DNA profiles.
Various methods can be used to formulate LR, including STRmix.
The likelihood ratio number could range up to 100 billion, and greater in certain jurisdiction.
The LR is a consideration of the likelihood of obtaining the observed DNA profile given two hypotheses, which should be the prosecution and defense propositions,
Valid DNA likelihood ratio calculation can turn on analyst inputs, including choice of propositions, allele frequency rate, theta correction, ethnicity considerations and relatives of the individual.
Contamination

DNA Contamination

DNA contamination can affect evidence from crime scene, laboratory and throughout the entire chain of custody.
Such contamination can impact upon the STRmix assessment.
Increased testing sensitivity means the majority of DNA profiles contain levels of background DNA. Background DNA will be analysed during the STRmix assessment and is commonly defined as an unknown contributor.
Background DNA of unknown contributors can represent contamination.
A rigorous review of the STRmix outputs, along with the laboratory contamination report can assist in the identification of any contamination that may have contributed to the STRmix assessment.
Error rate

STRmix Error Rate

STRmix error rate, known STRmix miscoding, and the variable nature of STRmix outputs, have been the subject of divisive scientific and legal debate through the evolution of STRmix and other PG (probabilistic genotyping) software development.
Multiple STRmix likelihood ratios will commonly be generated, and the analyst must exercise skill in the required subjective selection.
STRmix error reduction is reliant upon the appropriate formulation of propositions, contributor assigned and parameter settings, amongst other factors.
There are currently no published guidelines on the acceptable STRmix error rate.
Ethnicity database

DNA Ethnicity database

Valid DNA likelihood ratio calculation relies upon the appropriate ethnicity database selection.
In Australia, commonly selected datasets are Australian Aboriginal, Caucasian, Asian and Sudanese, though there are many others that can be utilised.
The numerical value (ie 100 billion) assigned to a DNA profile, requires an estimation as to how “common” or “rare” the profile is.
If the same DNA profile is assessed by STRmix using three different ethnicity databases’, it will produce three different likelihood ratios.
Certain laboratories have a tendency of applying a specific ethnicity, and others approach the matter on a case-by-case basis.
Number of contributors

Number of Contributors

A primary limitation of STRmix is that the analyst estimates the number of contributors.
Incorrect estimation of the number of contributors can directly cause an invalid likelihood ratio.
Contributor selection is critically reliant upon the DNA analyst’s skill, experience, and ability to evaluate factors of allele sharing, allele dropout, and stutter.
To minimise risk of bias, best practice state that the analyst conducts contributor selection without consideration to the person of interest samples.
Interrogation of the STRmix primary and secondary diagnostic outputs can assist to determine if the likelihood ratio is valid.
Helen Roebuck DNA expert Sydney

Why Roebuck Forensics

I have worked across every part of the forensic process - from crime scenes, to laboratory interpretation, to expert witness testimony.

My focus is not simply the presence of DNA or a statistic.

It is what the evidence can actually support — in the context of the case.

Next Step

If you’d like to discuss whether the forensic evidence in your matter warrants further review:

Contact Roebuck Forensics

A preliminary review can be a useful step in assessing the broad strengths and weaknesses of a particular matter. This process can also identify documentary and any further evidence requirements.

I rigorous interrogation of the evidence will uncover underlying issues and determine the most appropriate pathway towards reviewing and reporting the matter.

Certain complex matters benefit from a draft report, which can open areas for discussion with Counsel, and potentially allow for defence to approach the prosecution.

Following a thorough evaluation of the evidence, a DNA expert report will be issued in accordance with the expert witness code of conduct. The report will be suitable for submission in evidence.

Preparations with Counsel are often conducted, such that the probative value of the evidence is weighed effectively and persuasively at Voir Dire, should such a hearing be required.

Extensive preparations are generally conducted in anticipation of substantive hearing. Which may include cross examination and evidence in chief scenarios specific to the matter .

STRmix QUESTIONS

What is STRmix

Developed by the Institute of Environmental Science and Research (ESR), STRmix™ uses biological modeling and mathematical processes to resolve complex DNA mixtures. By utilising likelihood ratios (LRs) to quantify evidence for a hypothesis in a continuous approach, STRmix™ delivers faster, more accurate responses than binary methods previously in use.  STRmix was launched in 2012, and has produced legally admissible DNA evidence in more than 385,000 criminal cases worldwide. All Australian and New Zealand government laboratories use STRmix™ to analyse DNA evidence in criminal matters.

STRmix Likelihood Ratio

STRmix evaluates the DNA evidence generating a likelihood ratio (LR).  In simple terms, any LR greater than 1 provides support for the inclusion of the individual in question.  STRmix can also produce LRs supporting the exclusion of that individual from the DNA profile.

Who uses STRmix?

All Australian and New Zealand government laboratories use STRmix to analyse DNA evidence in criminal matters. STRmix™ is also being utilised in over 80 US laboratories in various regions, including Utah crime scene evidence and Alabama complex crime scene as well as in the United Kingdom and European facilities.

Access to STRmix

STRmix can be accessed by scientists engaged by the Defence.  Where neceesary we can conduct independent STRmix analyses, and further assess the results presented by the Prosecution.

STRmix false inclusions

When comparing a known non-donor DNA profile to the mixed DNA profile the STRmix software can sometimes produce a likelihood ratio supporting their contribution to the mixed DNA profile. In lay terms, STRmix™ can give a likelihood ratio that includes the accused as a contributor when in fact, they are not a contributor at all. This scenario is known as false inclusion.

 

False inclusion rates

STRmix false inclusion rates have been published in the STRmix developmental validation study, as well as other large scale studies.

 

Low likelihood ratios

The lower the likelihood ratio, the greater the risk of a false inclusion.

Proving a false inclusion

Determining whether a result is a false inclusion requires a skilled analyst.  Many factors must be considered, including the nature of the DNA profile, the number of contributors and the STRmix diagnostic results.

STRMix - Terms

Deconvolution

The process of separating a DNA profile into genotype sets and their associated probability weightings

Unified likelihood ratio

LR that takes into account that the unknown contributors within H1 (Hp) and H2 (Hd) are made of up of both relatives and unrelated people.

Stratified likelihood ratio

When multiple populations are selected to calculate an LR, STRmixTM will calculate LRs for each population individually and then provide a single LR that samples across all populations.

CONTACT
×

LAWYERS-DNA TOOLKIT

Enter