Home // Forensic // DNA Mixtures // Probabilistic Genotyping: Magic Hat or Mixture Resolution Tool?
Oct 17 2017
Probabilistic Genotyping: Magic Hat or Mixture Resolution Tool?
STRmix™ is an invaluable tool for resolving complex mixtures. Analysts are now able to make conclusions about samples that would previously have been reported as inconclusive, which in turn is leading to more crimes being solved. When the analysis is conducted properly, the results make sense based on the profile that we are analyzing with the software. However, it is not a magic hat where the analyst can throw any mixture into it, regardless of quality, and expect to pull out a great result. If a junky profile is entered into the software, the analyst will more than likely get an error or an uninformative likelihood ratio (LR) that is meaningless. In order to avoid this scenario, the mixed DNA profile should be evaluated prior to being imported into the software for analysis.
Written by: Cristina L. Rentas, DNA Labs International
At DNA Labs International, the STRmix™ software is not used for analysis of all of our samples. When possible, the traditional previously accepted methods of random match probability (RMP), modified RMP, or combined probability of inclusion (CPI) statistics that are still commonly used in the forensic DNA community today are employed. Once it is determined that a particular profile is inconclusive without using probabilistic genotyping methods, our analysts evaluate several aspects of the mixture in order to determine if a sample is eligible for additional analysis with probabilistic genotyping. Some of the aspects considered include; the number of contributors, how many loci have data and the overall quality of the profile.
When a profile is imported into the software, one of the determinations that the analyst needs to make is how many individuals contributed to that profile. Generally, the initial evaluation of determining potential number of contributors is done by counting the minimum number of alleles at each location. For example, you would expect there to be no more than two alleles at each locus per person, so if there are several locations that have five alleles, you would expect there to be at least three contributors to the mixture. Additionally, another one of the factors taken into consideration is disproportionate peak heights, which can also be an indication of an additional contributor. At our laboratory, we routinely analyze two, three, and four-person mixtures with the STRmix™ software. Depending on the quality of the profile, five-person mixtures may be run as well under certain circumstances. Generally, the five-person mixtures that have been run are from samples where the desired input target has been reached. It was found during our internal validation that five person mixtures that had been amplified with the optimal target produced more reliable results with the expanded kits than those that were amplified under target. This is due to the dropout that can be seen with samples that have a high number of contributors combined with a low quantity of input DNA.
In addition to allele count, the overall quality of the profile must be assessed as well. There must be enough data present in the profile to be able to conduct the analysis. While STRmix™ can account for dropout, highly degraded or inhibited samples may not provide enough information for the software and will likely result in an uninformative LR. In general, as the quantity of input DNA decreases, the LR does as well. This is especially true when the software is being used to evaluate low level minor DNA profiles. Once the mixture ratio becomes more extreme, it becomes more difficult to obtain a meaningful LR. However, it is important to note, that what was previously considered as extreme mixture proportions in DNA ten years ago is not the same as today. In the past, many laboratories would not work with mixtures greater than 1:10 or sometimes 1:30. With increased sensitivity in DNA testing, from improvements in extraction, amplification with expanded kits and capillary electrophoresis, we are now able to obtain meaningful data in mixture proportions out to 1:99.
- Teaching Old and New Dogs Mixture Tricks of the Future
- From Training to Casework: Validating and Implementing Probabilistic Genotyping in the Laboratory
After the profile has been evaluated and deemed suitable for import into STRmix™, comparisons can be made to any standards available for the case. Since we are reporting a LR and comparing two different scenarios, generally a standard is used for comparison. In general, if the majority of an individual’s types are not present in multiple loci where all contributors are represented in the mixture, our analysts are able to outright exclude that person without using the software. However, if many of the individual’s types are present and he cannot be ruled out as a contributor to the mixed DNA profile, then the software will be used and a final LR will be calculated. STRmix™ can also be used when no standard is available. If the software can deconvolute a contributor with a weight of over 99%, a value selected by our laboratory, then a profile is determined that can then be searched against a database or the mixture itself can be compared against a database for potential contributors to the mixture.
The most common sample type that is used for STRmix™ analysis at our laboratory is by far touch samples from guns. DNA samples obtained from swabbing guns typically result in low-level mixtures where a major and minor contributor cannot be determined. Often times these mixtures are ‘close’ to being suitable for comparison; i.e. allele sharing causes peak height ratios between the minor and major donor to be just above what we would comfortably use for RMP analysis; i.e. minor donor above approximately thirty-three percent of major donor for our laboratory. Each single analysis run occurs in two stages; the mixtures are deconvoluted and then after deconvolution, compared directly to a suspect. The ability to compare persons of interest to these mixtures allows for both exclusions and inclusions, which is essential for police investigations. Probabilistic genotyping is also ideal for sexual assault cases with multiple perpetrators that result in complex mixtures that previously couldn’t be evaluated. Furthermore, STRmix™ is a time-saver for the sexual assault backlog. Intimate sexual assault samples can processed in a ‘direct to DNA’ fashion without performing a differential extraction and subsequently the victim can be assumed, allowing the software to deduce the suspect profile. However, STRmix™ is not limited to these sample types; it can be used to analyze everything from a low-level single source sample obtained from a bottle to a complex mixture obtained from a rope. The analytical possibilities at our disposal with the addition of probabilistic genotyping as a tool in our arsenal are vast.
The introduction of probabilistic genotyping software into the forensic community has greatly changed forensic DNA analysis. Without this advancement, a number of cases would be halted and no further testing could be performed. However, it is always important for the analyst to remember that the software is not magic. It cannot make something appear when it is not there. By evaluating the profiles we upload and eliminating out these “junk” DNA samples, we can ensure that we are using the software as intended and obtaining the most accurate result possible.
Requests for additional information can be made via:
ATTN: STRmix inquiry
WOULD YOU LIKE TO SEE MORE ARTICLES LIKE THIS? SUBSCRIBE TO THE ISHI BLOG BELOW!