Completing the Full Workflow from Analysis to Interpretation to Data Matching
By Laura Russell, STRmix™ Ltd, Kenepuru Science Centre, Porirua, New Zealand
Share this article
While DNA evidence has been a staple of criminal investigations since the early 1990s, nothing since then has revolutionized forensic DNA interpretation like probabilistic genotyping (PG) software... and now forensic laboratories can do even more. Since its introduction more than a decade ago, PG software has enabled forensic labs to use far more of the available DNA profile to consider the probability of the evidence under two competing scenarios: typically, one that aligns with the prosecution standpoint (H1) and one that aligns with the defense (H2). As a result, low-level, degraded, or mixed DNA samples that previously would have been discarded as too complex have produced usable, interpretable, and legally admissible DNA evidence in more than half a million criminal cases worldwide.
Work by the same team that produced STRmix™, one of the first and most successful PG software packages, has led to the introduction of two complementary software applications: FaSTR™ DNA, which rapidly analyzes raw DNA data and assigns a number of contributors (NoC) estimate; and DBLR™, which allows forensic laboratories to visualize the value of DNA mixture evidence, carry out superfast database searches, and undertake extensive kinship analysis. When used in in combination with STRmix™, these applications deliver an end-to-end analysis, interpretation, and intelligence solution.
To illustrate this fully integrated workflow, let’s examine a hypothetical case in which swabs are taken from the grip of a handgun found at the scene of a crime for processing by the forensic laboratory.
Who could have donated any DNA detected? What is the profile quality like? Do we want to search all detected contributors or is there some risk to searching all? These are questions that may be addressed through use of the STRmix™ suite of software. Using the FaSTR™ DNA software, the laboratory can analyze data output by capillary electrophoresis (CE) instruments – the primary methodology used for separating and detecting short tandem repeat (STR) alleles in forensic DNA analysis.
FaSTR™ DNA analyzes the data applying dynamic baselining. Using peak detection and sizing algorithms, it may apply a set of laboratory-customized analysis rules to distinguish artefactual peaks from allelic peaks. It can also carry out number of contributors (NoC) estimation and sample-to-sample or sample-to-database comparison checks. Optionally, laboratories may employ an artificial neural network (ANN) to assist with profile interpretation. The exports and reports from FaSTR™ DNA are fully configurable.
In our hypothetical case, the FaSTR™ DNA analysis displays a mixed DNA profile from the swab taken from the grip of the handgun (Figure 1) and the smart decision-tree based NoC estimate indicates the DNA could originate from four individuals.
Figure 1: An excerpt from the FaSTR™ DNA electropherogram showing only information from the red and purple dyes of the DNA profile obtained from the swabs taken from the handgun.
Following FaSTR™ DNA analysis, results can be seamlessly exported straight to STRmix™, which uses sophisticated biological modelling and standard mathematical processes to interpret a wide range of complex DNA profiles.
Unlike previous methods of DNA analysis, which depended entirely on the application of fixed stochastic thresholds and other biological parameters to manually analyze DNA samples, STRmix™ uses a fully continuous model that makes better use of the information available within a DNA profile. It proposes hundreds of thousands of different profiles and is able to assess and weigh how closely these resemble the observed DNA mixture. STRmix™ relies on proven methodologies routinely used in computational biology, physics, engineering, and weather prediction and is able to use this information to assign a likelihood ratio (LR), the standard statistical approach to conveying weight of evidence information.
STRmix™ can interpret any kit or any number of re-amps, combine results generated from different kits, undertake statistical analysis on a range of contributor number (VarNOC), and customize the output.
The results of the STRmix™ analysis show mixture proportions for the four potential contributors of approximately 0.48, 0.42, 0.08, 0.02. From here, the laboratory can use DBLR™ to explore the discriminatory power of the different mixture components and assess the risk/benefit of comparing to reference profiles (standards) or searching against a database.
DBLR™ uses efficient algorithms for rapidly calculating LRs. In addition to the Explore Deconvolution and Search Database modules, the DBLR™ Kinship function enables a laboratory to load STRmix™ deconvolutions or single-source profiles from known individuals and link these with one or more pedigrees. The Mixture to Mixture functionality helps to address questions such as whether two mixtures have a common source, while the Common Donor functionality considers who a common donor may be. Using Explore Deconvolution module, DBLR™ is able to simulate thousands of profiles in a fraction of a second for putative true contributors (H1 true), based on the STRmix™ weights, and non-contributor profiles (H2 true) by sampling from the selected allele frequency database.
The software plots distributions for the expected range of H1 true LRs (shown in blue) and H2 LRs (shown in red).
In our hypothetical case (Figure 2), component 3 of this mixture appears suitable for comparison. Based on the distributions of expected LRs, we may expect good discrimination capacity and should be able to clearly distinguish true from false donors. The Search Database function of DBLR™ can then be used to search component 3 against saved databases in seconds. This comparison leads to a link to an individual on the database with an assigned log10(LR) of 17.4. Component 4, on the other hand, displays significant risk of false exclusion or false inclusion. A decision not to progress the interpretation of this component is made based on these results.
Bottom line, the combination of FaSTR™ DNA plus STRmix™ plus DBLR™ provides a total integrated workflow solution, turning raw data into crucial evaluative and intelligence information for investigators.
Figure 2: Distributions displaying the range of expected LRs for H1 true (blue) and H2 True (red) for component 3 (top pane) and component 4 (bottom pane) of the mixed DNA profile obtained from swabs taken from the grip of a handgun.
Reference/Acknowledgement
The mixed DNA profile used in this publication was created as part of the data set described in: L.E. Alfonse, A.D. Garrett, D.S. Lun, K.R. Duffy, C.M. Grgicak, A large-scale dataset of single and mixed-source short tandem repeat profiles to inform human identification strategies: PROVEDIt, Forensic Sci. Int. Genet. 32 (2018) 62-70.
Image header: AdobeStock/Who is Danny
Laura Russell
Senior Scientist on the STRmix™ Ltd, Porirua, New Zealand
Laura Russell is a Senior Scientist on the STRmix team. A 20-year veteran in the forensic field, Russell has casework experience in both New Zealand and the UK. She currently provides training and support to users of STRmix™ software, while also serving as a Quality Officer.
For more information, visit www.strmix.com.