Finding the way forward for forensic science in the US-A commentary on the PCAST report.

A recent report by the US President's Council of Advisors on Science and Technology (PCAST), (2016) has made a number of recommendations for the future development of forensic science. Whereas we all agree that there is much need for change, we find that the PCAST report recommendations are founded on serious misunderstandings. We explain the traditional forensic paradigms of match and identification and the more recent foundation of the logical approach to evidence evaluation. This forms the groundwork for exposing many sources of confusion in the PCAST report. We explain how the notion of treating the scientist as a black box and the assignment of evidential weight through error rates is overly restrictive and misconceived. Our own view sees inferential logic, the development of calibrated knowledge and understanding of scientists as the core of the advance of the profession.

DNA Commission of the International Society for Forensic Genetics: Recommendations on the validation of software programs performing biostatistical calculations for forensic genetics applications.

The use of biostatistical software programs to assist in data interpretation and calculate likelihood ratios is essential to forensic geneticists and part of the daily case work flow for both kinship and DNA identification laboratories. Previous recommendations issued by the DNA Commission of the International Society for Forensic Genetics (ISFG) covered the application of bio-statistical evaluations for STR typing results in identification and kinship cases, and this is now being expanded to provide best practices regarding validation and verification of the software required for these calculations. With larger multiplexes, more complex mixtures, and increasing requests for extended family testing, laboratories are relying more than ever on specific software solutions and sufficient validation, training and extensive documentation are of upmost importance. Here, we present recommendations for the minimum requirements to validate bio-statistical software to be used in forensic genetics. We distinguish between developmental validation and the responsibilities of the software developer or provider, and the internal validation studies to be performed by the end user. Recommendations for the software provider address, for example, the documentation of the underlying models used by the software, validation data expectations, version control, implementation and training support, as well as continuity and user notifications. For the internal validations the recommendations include: creating a validation plan, requirements for the range of samples to be tested, Standard Operating Procedure development, and internal laboratory training and education. To ensure that all laboratories have access to a wide range of samples for validation and training purposes the ISFG DNA commission encourages collaborative studies and public repositories of STR typing results.

An illustration of the effect of various sources of uncertainty on DNA likelihood ratio calculations.

A typical assessment of the strength of forensic DNA evidence is based on a population genetic model and estimated allele frequencies determined from a population database. Some experts provide a confidence or credible interval which takes into account the sampling variation inherent in deriving these estimates from only a sample of a total population. This interval is given in conjunction with the statistic of interest, be it a likelihood ratio (LR), match probability, or cumulative probability of inclusion. Bayesian methods of addressing database sampling variation produce a distribution for the statistic from which the bound(s) of the desired interval can be determined. Population database sampling uncertainty represents only one of the sources of uncertainty that affects estimation of the strength of DNA evidence. There are other uncertainties which can potentially have a much larger effect on the statistic such as, those inherent in the value of Fst, the weights given to genotype combinations in a continuous interpretation model, and the composition of the relevant population. In this paper we model the effect of each of these sources of uncertainty on a likelihood ratio (LR) calculation and demonstrate how changes in the distribution of these parameters affect the reported value. In addition, we illustrate the impact the different approaches of accounting for sampling uncertainties has on the LR for a four person mixture.

The interpretation of lineage markers in forensic DNA testing.

Mitochondrial DNA (mtDNA) and the non-recombining portion of the Y-chromosome are inherited matrilinealy and patrilinealy, respectively, and without recombination. Collectively they are termed 'lineage markers'. Lineage markers may be used in forensic testing of an item, such as a hair from a crime scene, against a hypothesised source, or in relationship testing. An estimate of the evidential weight of a match is usually provided by a count of the occurrence in some database of the mtDNA or Y-STR haplotype under consideration. When the factual statement of a count in the database is applied to a case, issues of relevance of the database and sampling uncertainty may arise. In this paper, we re-examine the issues of sampling uncertainty, the relevance of the database, and the combination of autosomal and lineage marker evidence. We also review the recent developments by C.H. Brenner.

Use of DNA profiles for investigation using a simulated national DNA database: Part I. Partial SGM Plus profiles.

In traditional criminal investigation, uncertainties are often dealt with using a combination of common sense, practical considerations and experience, but rarely with tailored statistical models. For example, in some countries, in order to search for a given profile in the national DNA database, it must have allelic information for six or more of the ten SGM Plus loci for a simple trace. If the profile does not have this amount of information then it cannot be searched in the national DNA database (NDNAD). This requirement (of a result at six or more loci) is not based on a statistical approach, but rather on the feeling that six or more would be sufficient. A statistical approach, however, could be more rigorous and objective and would take into consideration factors such as the probability of adventitious matches relative to the actual database size and/or investigator's requirements in a sensible way. Therefore, this research was undertaken to establish scientific foundations pertaining to the use of partial SGM Plus loci profiles (or similar) for investigation.

Use of DNA profiles for investigation using a simulated national DNA database: Part II. Statistical and ethical considerations on familial searching.

Familial searching consists of searching for a full profile left at a crime scene in a National DNA Database (NDNAD). In this paper we are interested in the circumstance where no full match is returned, but a partial match is found between a database member's profile and the crime stain. Because close relatives share more of their DNA than unrelated persons, this partial match may indicate that the crime stain was left by a close relative of the person with whom the partial match was found. This approach has successfully solved important crimes in the UK and the USA. In a previous paper, a model, which takes into account substructure and siblings, was used to simulate a NDNAD. In this paper, we have used this model to test the usefulness of familial searching and offer guidelines for pre-assessment of the cases based on the likelihood ratio. Siblings of "persons" present in the simulated Swiss NDNAD were created. These profiles (N=10,000) were used as traces and were then compared to the whole database (N=100,000). The statistical results obtained show that the technique has great potential confirming the findings of previous studies. However, effectiveness of the technique is only one part of the story. Familial searching has juridical and ethical aspects that should not be ignored. In Switzerland for example, there are no specific guidelines to the legality or otherwise of familial searching. This article both presents statistical results, and addresses criminological and civil liberties aspects to take into account risks and benefits of familial searching.

An investigation into the relationship between edge counts and the variability of the refractive index of glass. Part I: Edge morphology.

The effect of glass fragment morphology on the variability of refractive index values in five different glass objects was investigated. Data are presented that suggest that the variability of refractive index values is increased when fragment edge morphology becomes unsuitable for phase contrast microscopy.

An extended likelihood ratio framework for interpreting evidence.

This paper reviews some current methods, the likelihood ratio-based approach and the full Bayesian approach for the interpretation of evidence and discusses previously identified shortcomings in them. It suggests an approach based on a compromise--based on an extended likelihood ratio--that may combine the merits of logic without overstepping acceptable bounds for the forensic scientist in the presentation of evidence. The approach is exposed formally and takes advantage of inferential networks called Bayesian networks.

DNA commission of the International Society of Forensic Genetics: Recommendations on the interpretation of mixtures.

The DNA commission of the International Society of Forensic Genetics (ISFG) was convened at the 21st congress of the International Society for Forensic Genetics held between 13 and 17 September in the Azores, Portugal. The purpose of the group was to agree on guidelines to encourage best practice that can be universally applied to assist with mixture interpretation. In addition the commission was tasked to provide guidance on low copy number (LCN) reporting. Our discussions have highlighted a significant need for continuing education and research into this area. We have attempted to present a consensus from experts but to be practical we do not claim to have conveyed a clear vision in every respect in this difficult subject. For this reason, we propose to allow a period of time for feedback and reflection by the scientific community. Then the DNA commission will meet again to consider further recommendations.

A study of the performance and utility of annealing in forensic glass analysis.

The use of annealing in forensic glass casework is reviewed. New data is presented that suggests a continuous approach to the interpretation of such data may be superior to the classification approach. Data are presented supporting the previously observed relationship between change in refractive index (RI) and the thickness of the glass pane. An examination of the possible assumption of independence between RI and change in RI is undertaken.

The two trace transfer problem re-examined.

The classical two trace transfer problem initially discussed by Evett is reconsidered [Evett IW. On Meaningful Questions: A Two-Trace Transfer Problem. Journal of the Forensic Science Society 1987; 27: 375-381]. In this problem we consider the situation where there are two stains of differing types at a crime scene. Evett considered the situation where both are known to be relevant to the offence. It was further assumed that there were exactly two offenders. Reanalysis of this problem suggests that the key consideration is the conditioning on there being exactly two offenders, not that there were two stains. The treatment here leads easily to a generalisation in which stains have differing relevance and there are k offenders.

Effects of population structure and admixture on exact tests for association between Loci.

The probability of multilocus genotype counts conditional on allelic counts and on allelic independence provides a test statistic for independence within and between loci. As the number of loci increases and each sampled genotype becomes unique, the conditional probability becomes a function of total heterozygosity. In that case, it does not address between-locus dependence directly but only indirectly through detection of the Wahlund effect. Moreover, the test will reject the hypothesis of allelic independence only for small values of heterozygosity. Low heterozygosity is expected for population subdivision but not for population admixture. The test may therefore be inappropriate for admixed populations. If individuals with parents in two different populations are always considered to belong to one of the populations, then heterozygosity is increased in that population and the exact test should not be used for sparse data sets from that population. If such a case is suspected, then alternative testing strategies are suggested.

Allele frequencies for the four major sub-populations of New Zealand at three STR loci--HUMTH01, HUMTPOX and CSF1PO.

Allele frequencies for the three STR loci included in the GenePrint CTT multiplex system (HUMTH01, HUMTPOX, HUMCSF1PO) have been determined for the four major sub-populations of New Zealand.

Assessing uncertainty in DNA evidence caused by sampling effects.

Sampling error estimation in forensic DNA testimony was discussed. Is an estimate necessary and how should it be made? The authors find that all modern methods have areas of strength and weakness. The assessment of which is the 'best' is subjective and depends on the performance of the method, the type of problem (criminal work or paternity), the database size and availability of computing software and support. The authors preferred the highest posterior density approach for performance, however the other methods all have areas where their performance is adequate. For single-contributor stains normal approximation methods are suitable, also the bootstrap and the highest posterior density method. For multiple-contributor stains or other complex situations the match probability expressions become quite complex and it may not be possible to derive the necessary variance expressions. The highest posterior density or the bootstrap provide a better general method, with non-zero theta. The size-bias correction and the factor of 10 approaches may be considered acceptable by many forensic scientists as long as their limitations are understood.

How many samples from a drug seizure need to be analyzed?

Recently Aitken introduced an outstanding advance in the approach to decision making regarding drugs sampling. Unfortunately this approach has not, as yet, been widely implemented despite being based on a solid mathematical foundation. In this paper we advocate a Bayesian approach along the lines of that outlined by Aitken but designed to be both easily understood with less mathematical sophistication and implementable using standard EXCEL software. The emphasis is placed on encouraging the application of this methodology to routine case work by explaining the statistics involved. Minor differences exist between this approach and that of Aitken in both the modeling of the prior probability and in dealing with the discrete nature of the samples. These differences in no way detract from the sound mathematical foundation of the approach.

The fallacy of independence testing and the use of the product rule.

Use of the product rule which implies the assumption of within and between locus independence, is still common, particularly in the United States of America. Whilst it may be considered by some to be an acceptable approximation it is not logical to suggest that independence testing somehow "validates" its use. This paper discusses the nature of this fallacy.

Glass on clothing and shoes of members of the general population and people suspected of breaking crimes.

The outer clothing and footwear of 122 people attending a university gymnasium and a private gymnasium were searched for fragments of glass. Both the surfaces and the pockets of the clothing and the uppers and soles of the footwear were searched. New Zealand forensic glass cases have been reviewed to determine the amount of non-matching glass present on the clothing of people who are suspected of breaking crimes. Data from 114 suspects who had no matching glass on their clothing and shoes were accumulated. Statistical modelling techniques have been applied to the data collected.

An investigation of the rigor of interpretation rules for STRs derived from less than 100 pg of DNA.

By increasing the PCR amplification regime to 34 cycles, we have demonstrated that it is possible routinely to analyse <100 pg DNA. The success rate was not improved (without impairing quality) by increasing cycle number further. Compared to amplification of 1 ng DNA at 28 cycles, it was shown that increased imbalance of heterozygotes occurred, along with an increase in the size (peak area) of stutters. The analysis of mixtures by peak area measurement becomes increasingly difficult as the sample size is reduced. Laboratory-based contamination cannot be completely avoided, even when analysis is carried out under stringent conditions of cleanliness. A set of guidelines that utilises duplication of results to interpret profiles originating from picogram levels of DNA is introduced. We demonstrate that the duplication guideline is robust by applying a statistical theory that models three key parameters - namely the incidence of allele drop-out, laboratory contamination and stutter. The advantage of the model is that the critical levels for each parameter can be calculated. This information may be used (for example) to determine levels of contamination that can be tolerated within the strategy employed. In addition we demonstrate that interpreting one banded loci, where allele dropout could have occurred, using LR=1/2f(a) was conservative provided that the band was low in peak area. Furthermore, we demonstrate that an apparent mis-match between crime-stain and a suspect DNA profile does not necessarily result in an exclusion. The method used is complex, yet can be converted into an expert system. We envisage this to be the next step.

Evaluating the statistical significance of single band profiles in VNTR analyses.

VNTR profiles may present either a single band or two bands. If two bands are present then the individual is a heterozygote for these two bands. However, if only one band is present there is ambiguity as to the true genotype of the individual. This person may be a homozygote in that he has two copies of the same allele, or he may be a heterozygote for two very close bands that cannot be separated on the gel. The second NRC report proposed the use of the '2p' rule, or Formula 4.10a in the sub-structure case, as a conservative upper bound in the statistical interpretation. However, further examination suggests that these formulae are not necessarily conservative. In this paper we examine this phenomenon by deriving a formula that contains both the corrections for null alleles and for subpopulation effects.