Use of the LUS in sequence allele designations to facilitate probabilistic genotyping of NGS-based STR typing results

Abstract

Some of the expected advantages of next generation sequencing (NGS) for short tandem repeat (STR) typing include enhanced mixture detection and genotype resolution via sequence variation among non-homologous alleles of the same length. However, at the same time that NGS methods for forensic DNA typing have advanced in recent years, many caseworking laboratories have implemented or are transitioning to probabilistic genotyping to assist the interpretation of complex autosomal STR typing results. Current probabilistic software programs are designed for length-based data, and were not intended to accommodate sequence strings as the product input. Yet to leverage the benefits of NGS for enhanced genotyping and mixture deconvolution, the sequence variation among same-length products must be utilized in some form. Here, we propose use of the longest uninterrupted stretch (LUS) in allele designations as a simple method to represent sequence variation within the STR repeat regions and facilitate - in the nearterm - probabilistic interpretation of NGS-based typing results. An examination of published population data indicated that a reference LUS region is straightforward to define for most autosomal STR loci, and that using repeat unit plus LUS length as the allele designator can represent greater than 80% of the alleles detected by sequencing. A proof of concept study performed using a freely available probabilistic software demonstrated that the LUS length can be used in allele designations when a program does not require alleles to be integers, and that utilizing sequence information improves interpretation of both single-source and mixed contributor STR typing results as compared to using repeat unit information alone. The LUS concept for allele designation maintains the repeat-based allele nomenclature that will permit backward compatibility to extant STR databases, and the LUS lengths themselves will be concordant regardless of the NGS assay or analysis tools employed. Further, these biologically based, easy-to-derive designations uphold clear relationships between parent alleles and their stutter products, enabling analysis in fully continuous probabilistic programs that model stutter while avoiding the algorithmic complexities that come with string based searches. Though using repeat unit plus LUS length as the allele designator does not capture variation that occurs outside of the core repeat regions, this straightforward approach would permit the large majority of known STR sequence variation to be used for mixture deconvolution and, in turn, result in more informative mixture statistics in the near term. Ultimately, the method could bridge the gap from current length-based probabilistic systems to facilitate broader adoption of NGS by forensic DNA testing laboratories.

Systematic evaluation of the early access applied biosystems precision ID Globalfiler mixture ID and Globalfiler NGS STR panels for the ion S5 system

“…In line with our findings, also Wang et al observed higher stutter ratios for D22S1045, D1S1656 and D12S391 using the GF-STR panel [21]. As it is known that stutter ratio depends on the sequence of the corresponding STR allele [49], the sequence effect on stutter height was analzsed by GMI exemplarily for D12S391 (repeat structure: Fig. 1.…”

Identification of sequence polymorphisms at 58 STRs and 94 iiSNPs in a Tibetan population using massively parallel sequencing

“…It is urgent to know how the MPS data should be analysed and reported, what connections do these data have with LB alleles, and how to record and search such datasets in a database 4 . Some researchers have tried to answer these questions 19,20 but a perfect nomenclature is still under development. A unified minimal nomenclature of the complex sequences obtained by MPS technologies was recommended by the International Society for Forensic Genetics (ISFG) in 2016 21,22 to facilitate communication between laboratories and to make this data backward compatible with LB data produced on CE platform.…”

Automation of high volume MPS mixture interpretation using CaseSolver

“…When LR (log10) was inconclusive (-2 to 2) for RU, use of the LUS changed the LR (log10) by 0.3 on average, while from LUS to LUS+ the average LR change was only -0.02. When the RU LR (log10) provided some support for inclusion (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12), the increase in LR (log10) by use of LUS averaged 1.2 (3.6 if any support), and when the LUS LR provided any support for inclusion (>2), whereas the increase in LR (log10) by use of LUS+ averaged 0.6. Similarly, when starting LR (log10) values were in the range of 2-12, the LR increased by 2+ orders of magnitude more frequently when LUS was compared to RU interpretations (31%) versus when LUS+ was compared to LUS interpretations (4%).…”

“…Here the number of points for both RU->LUS and LUS→LUS+ was one. (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12) and provide very strong support for inclusion (>12). The green triangle indicates some support for inclusion initially (RU or LUS) but provided very strong support for inclusion when more sequence information (LUS or LUS+) was used.…”

“…Here, we provide a demonstration using 60 mixtures typed using the ForenSeq DNA Signature Prep kit (Verogen, Inc., San Diego, CA). A comparison of results using different levels of sequence information was carried out using ordinary repeat units (RU), the longest uninterrupted stretch (LUS) representations [2], and LUS+ representations [3] that captured all sequence variation present in the donor genotypes. We report preliminary results on the extent of enhanced discrimination offered by MPS when complex mixtures are analysed.…”