Perlin Mark W, Legler Matthew M, Spencer Cara E, Smith Jessica L, Allan William P, Belrose Jamie L, Duceman Barry W
Cybergenetics Corp, 160 North Craig Street, Suite 210, Pittsburgh, PA 15213, USA.
J Forensic Sci. 2011 Nov;56(6):1430-47. doi: 10.1111/j.1556-4029.2011.01859.x. Epub 2011 Aug 9.
DNA mixtures with two or more contributors are a prevalent form of biological evidence. Mixture interpretation is complicated by the possibility of different genotype combinations that can explain the short tandem repeat (STR) data. Current human review simplifies this interpretation by applying thresholds to qualitatively treat STR data peaks as all-or-none events and assigning allele pairs equal likelihood. Computer review, however, can work instead with all the quantitative data to preserve more identification information. The present study examined the extent to which quantitative computer interpretation could elicit more identification information than human review from the same adjudicated two-person mixture data. The base 10 logarithm of a DNA match statistic is a standard information measure that permits such a comparison. On eight mixtures having two unknown contributors, we found that quantitative computer interpretation gave an average information increase of 6.24 log units (min = 2.32, max = 10.49) over qualitative human review. On eight other mixtures with a known victim reference and one unknown contributor, quantitative interpretation averaged a 4.67 log factor increase (min = 1.00, max = 11.31) over qualitative review. This study provides a general treatment of DNA interpretation methods (including mixtures) that encompasses both quantitative and qualitative review. Validation methods are introduced that can assess the efficacy and reproducibility of any DNA interpretation method. An in-depth case example highlights 10 reasons (at 10 different loci) why quantitative probability modeling preserves more identification information than qualitative threshold methods. The results validate TrueAllele(®) DNA mixture interpretation and establish a significant information improvement over human review.
含有两个或更多贡献者的DNA混合样本是生物证据的一种普遍形式。由于存在不同基因型组合的可能性，这些组合可以解释短串联重复序列（STR）数据，使得混合样本的解读变得复杂。目前人工审查通过应用阈值将STR数据峰定性地视为全有或全无事件，并赋予等位基因对相等的可能性，从而简化了这种解读。然而，计算机审查可以处理所有定量数据，以保留更多识别信息。本研究考察了在相同的经裁决的两人混合样本数据中，与人工审查相比，定量计算机解读在多大程度上能够获取更多识别信息。DNA匹配统计量的常用对数是一种标准信息度量，它允许进行这种比较。在八个有两个未知贡献者的混合样本上，我们发现与定性人工审查相比，定量计算机解读平均信息增加了6.24个对数单位（最小值 = 2.32，最大值 = 10.49）。在另外八个有已知受害者参考样本和一个未知贡献者的混合样本上，与定性审查相比，定量解读平均增加了4.67个对数因子（最小值 = 1.00，最大值 = 11.31）。本研究对DNA解读方法（包括混合样本）进行了全面探讨，涵盖了定量和定性审查。引入了验证方法，可评估任何DNA解读方法的有效性和可重复性。一个深入的案例示例突出了10个原因（在10个不同位点），说明为什么定量概率建模比定性阈值方法保留了更多识别信息。结果验证了TrueAllele(®) DNA混合样本解读，并确立了与人工审查相比显著的信息提升。
