Evaluating a peak height based method to determine the number of contributors in a DNA mixture and a study of DNA recovery using laser microdissection

Forensic laboratories process evidentiary samples found at crime scenes. These evidentiary samples may contain limited amounts of DNA and DNA from more than one individual. When data from these sample types are acquired, the data may be so complicated that efficient interpretation is prohibitive. Thus, there is a need for optimal DNA processing during testing. The forensic analysis of DNA involves; extraction, to purify the DNA; quantification, to determine the amount of DNA; amplification, to replicate DNA fragments of interest; separation and detection of the DNA fragments; and interpretation of genetic profiles. With each process there is error introduced, such as pipetting error and stochastic effects, which introduce the need to consider stutter artifacts and preferential amplification during interpretation. The result is difficult and complex profile analysis and interpretation. Since the work performed in a forensic laboratory involves human identification for purposes of resolving a legal argument, it is of importance to minimize error and obtain the most accurate and precise result for comparison. Thus, the first part of this study explores recovery rates for an alternative sample preparation method: laser microdissection. Laser microdissection is proclaimed to be less time consuming compared to other methodologies, but more expensive; and many crime laboratories may deem the investment in such a system prohibitive. In this study, the majority of the DNA recovery rates of LMD samples were less than 10%, which is much lower than previously published recovery rates of 16-32%. Therefore, further studies optimizing the LMD method should be performed prior to the implementation in forensic casework. The second part of this study assessed the possibility of using average peak height in lieu of template mass when evaluating the number of contributors. The evaluation was done using the probabilistic software NOCIt, which is a computational tool created to aid analysts in determining the likely number of contributors (i.e. NOC) to a sample. The software currently uses target mass as a basis for its models and calculations. However, it has been suggested that target mass may be a suboptimal explanatory variable. In an effort to ascertain whether the use of target mass negatively impacts results, average peak height was analyzed to examine whether it could be used as the independent variable within NOCIt and to discern whether the results were comparable to those obtained when target mass was the independent variable. Prior to use of average peak height in NOCIt, an examination of the models that characterize alleles, noise, allele drop-out, reverse and forward stutter, and reverse and forward stutter drop-out was conducted. This study demonstrated concordance between the class of models which use target mass and the models that use average peak height, indicating that average peak height could be used as the independent variable in NOCIt. Further, since NOCIt models both noise and forward stutter, an evaluation of the effects of characterizing forward stutter as noise ensued. The results show that at low target masses the effects of characterizing forward stutter as noise are negligible; however an effect is seen at higher template mass. Thus, the model which does not characterize forward stutter as noise was preferred since it can be utilized for both low-template and high-template masses. Prior to the assessment of NOCIt utilizing APH (average peak height), an examination of the software reproducibility using TM (target mass) was performed to ensure that multiple runs would estimate the same NOC with a similar a posteriori probability (APP). For this study, 128 samples were analyzed in quintuplicates; and, only 12 samples showed a different estimate for NOC. A repeatability graph plotting the range of APP against the median APP for quintuplicate samples illustrated that when the median APP was high (approximately 0.999), the range of APP was small (approximately 0.1). The data suggest that as the median APP decreases and the range of APP increases, there is less of a chance that the actual NOC can be determined with certainty. After repeatability results were determined, the use of the APH as the independent variable was analyzed. The accuracies were compared to those obtained when TM was used. If the max probability was taken as an indicator of NOC, the accuracy was 48 % and 47 % for APH and TM, respectively, illustrating that APH could be used by NOCIt as a proxy for TM