Statement on minor miscoding in BN Formulae STRmix 1.08 with additional comments. Statement relating to STRmix miscodes

Transcription

1 During the lifetime of the project we have detected seven coding faults that act on the reported LR. In no case was an inclusion changed to an exclusion or vice versa. 1. Present in versions up to, but not including, It affected the LR in an exceptionally minor way and produced neither false inclusions nor exclusions. It was detected by a third party laboratory repeating calculations by hand. It did not result in the need to reissue any statements. Document Statement on minor miscoding in BN Formulae STRmix 1.08 with additional comments. 2. Present in STRmix versions up to but not including version This was discovered after a case was brought to Forensic Science South Australia s attention in December 2014 by Queensland Health. This miscode was reported, with errors in the reporting, in the press. 1 The miscode affects the numerical value of the LR in rare instances and does not change an inclusion into an exclusion. All potentially affected cases in Australia and New Zealand were examined and very few instances of a different LR occurred outside Queensland. There are no known false inclusions as a result of the miscode. To get an effect from this miscode a specific set of circumstances is needed which we believed quite unlikely to occur and which had not been specifically tested in developmental validation prior to this date. Document: Statement relating to STRmix miscodes Involved rare cases with multiple dropped alleles across all contributors within the one locus. This was found to be due to a change in STRmix that was introduced in the V2.3 series and has been present in all versions subsequent to this up to V and V (inclusive). The effect on likelihood ratios (LRs) is in a conservative direction and very minor. Document: FINAL report - variation in STRmix regarding calculation of expected heights of dropped out peaks, 4 July Present in V1.0 to V for drop-in modelling and V2.4 to V for forward stutter modelling. This affects only the database search LR (assuming the presence of forward stutter peaks and/or multiple drop-in alleles, and modelling of these artifacts was enabled) and to the Highest Posterior Density (HPD) calculations. There was no detectable effect on the LR in a number of trials. STRmix v Release and Testing Report authorities confirm miscode affects dnaevidence in criminal cases/news story/833c580d3f1c59039efd1a2ef55af92b The number of cases affected is 23 not 60 see data/assets/pdf_file/0029/633368/dohdl pdf We do not have good data on the details of these cases. Queensland Health shared data up to and including the first of the set of profiles associated with this, then ceased citing initially that sharing of profiles was now illegal. Hence we do not know the magnitude of any change. Subsequently Queensland Health have informed us that they can share the profiles but this has never occurred. We speculate that the much higher rate of exposure of this miscode in Queensland is a result of their workflow which, we believe, in some instances adds an extra contributor beyond that required at a minimum to explain the profile. This exposed a part of the code with an error in it. This usage had not been envisaged and had not be tested prior to this event.

2 5. Present in all versions preceding v A change has been made to LR calculation (unrelated point estimate, stratified, unified and HPD) for mixed DNA profiles when there are multiple contributors considered under Hp who are unknown under Hd. The contributors must have dropped alleles (either the same or different) at the same locus. Changes were usually less than a factor of 10. See appendix 6. Present in all versions preceding v A change has been made to the way drop-in alleles are assigned within the determination of the genotype array within pre burn-in. This will affect all LR calculations including within the database search (standard and familial LRs), unrelated and related scenarios. Changes in the LR were usually less than one order of magnitude. See appendix 7. Present in all versions preceding v A minor anomaly in the familial search LR was identified. The issue affected mixed and single source profiles with dropout where on some occasions an incorrect allele frequency was assigned to the alleles. The comparison of LRs between versions indicates that this was mostly within one order of magnitude. See appendix

3 Appendix 13 June 2017 <snip> Summary of changes: Three changes have been made to the likelihood ratio (LR) calculations in STRmix resulting in v This updates the LR calculations in all versions preceding v A change has been made to the LR calculation (unrelated point estimate, stratified, unified and HPD) for mixed DNA profiles when there are multiple contributors considered under Hp who are unknown under Hd. The contributors must each have dropped alleles (of either the same or different types) at the same locus. Examples of affected propositions include: Hp1: DNA has originated from POI 1 and POI 2 Hd1: DNA has originated from two unknown individuals Hp2: DNA has originated from POI 1, POI 2 and POI 3 Hd2: DNA has originated from three unknown individuals Hp3: DNA has originated from POI 1, POI 2 and POI 3 Hd3: DNA has originated from POI 1 and two unknown individuals This change relates to the order of the contributors within the LR calculation. The magnitude of the effect on LR calculations will depend on the proposition set used within the calculation and the number of per locus LR calculations undertaken where both contributors are dropping out of the mixture at the same locus. The effect is reduced when the number of loci of this specific configuration is reduced and when the difference in unknowns between Hp and Hd is reduced. 85 two, three and four person low level mixed DNA profile interpretations were carried out in STRmix v and v in order to size the difference in LR. The log(lr) is compared in the plot below. The dashed line is at x=y. The dotted lines represented ± one order of magnitude. Different propositions were considered with varying numbers of contributors assumed under Hp (two, three or four). Forty calculations out of the 85 tested resulted in different LRs. The difference in all bar one

4 example was within one order of magnitude. The largest difference was an LR=114 using v and LR=0.2 using v This was a known four person mixture interpreted assuming four contributors, analysed under conditions that were deliberately assigned to maximise the possible difference in LRs. There were a maximum of four peaks at any one locus and the maximum allele height was 225 rfu. The profile had previously been designated as an apparent three person mixture. Three of the four known contributors were assumed under Hp. 2. A change has been made to the way drop-in alleles are assigned during the determination of the genotype array within pre burn-in. This will affect all LR calculations including the database search module (standard and familial LRs), unrelated and related scenarios. Within v a peak in a double back stutter position (peak a in the figure above) of an allele (peak c) that is less

5 than the maximum back stutter parameter (percent) of the stutter peak (peak b) was assigned as stutter of stutter. Within v this peak is now modelled as either allelic or drop-in. This change will only affect interpretations where drop-in is modelled ( turned on ) within the kit. Within single source and mixed DNA profiles where the peak is solely modelled as drop-in, the drop-in penalty cancels under Hp and Hd and the LR will be numerically identical between versions. For profiles where the peak is modelled as both allelic and drop-in under various genotype combinations a difference in LR between versions is expected. A number of twoperson profiles were edited by adding a peak in a double back stutter position of a major contributor s allele. The minor contributor was of a similar height to the added peak to allow consideration of the peak as both allelic and drop-in. A comparison of the log(lr) calculated using v and v is given in the plot below. The dashed line is at x=y. The dotted lines represented ± one order of magnitude. The comparison of LRs between versions indicates that this was predominantly within one order of magnitude. 3. A minor anomaly in the familial search LR was identified. The issue affected mixed and single source profiles with dropout where on some occasions an incorrect allele frequency was assigned to the alleles. A comparison of familial LRs calculated using the database search functionality between versions is plotted below where the dashed line is at x=y. The dotted lines represented ± one order of magnitude. The comparison of LRs between versions indicates that

6 this was mostly within one order of magnitude. This comparison included profiles with drop-in, dropout, and a combination of both drop-in and dropout modelled within the interpretation, with up to 4 contributors. The familial LRs were calculated for both known contributors and noncontributors to the profiles. There is more variability in the LR where more dropout is observed (approximately less than LR = 1000).