Characterizing rates of allelic dropout and the impact on estimating the number of contributors

Forensic analysis of a deoxyribonucleic acid (DNA) profile includes determining if DNA from a known person should be considered as a likely contributor to the biological evidence. Prior to making this determination, the number of contributors (NOC) of the DNA sample is considered. It is important to take multiple factors into account when estimating the NOC, including stutter, baseline noise, and peak imbalance as these can affect the number of peaks observed at each locus. Allelic dropout can also have an impact on the number of peaks observed. Dropout occurs when an allele is not detected due to technical, biochemical, or sampling issues, and predominantly affects the level of ambiguity associated with low-template DNA interpretation. As the NOC to a sample may be underestimated in the presence of dropout, it is essential to reasonably predict the probability that an allele has dropped out. This work has two aims: first, to evaluate different characterizations of allelic dropout rates and, second, to determine the impact allelic dropout has on estimating the NOC to a DNA sample. Two different types of dropout characterization were examined – ‘indirect’ models based on observed peak heights and ‘direct’ models using observed dropout frequencies of single-source calibration data. The indirect models predicted allelic dropout based on the peak height distribution of the data at a specific target amount and locus using a fitted or non-fitted cumulative Gaussian curve. For the direct models, a logistic or exponential regression of the observed dropout frequencies versus target amount for each locus was used to predict dropout rates. The impact that allelic dropout has on estimating the NOC was assessed by varying the probability of dropout (Pr(D)) in simulated mixtures with up to six contributors in the presence or absence of a major contributor. Simulations for the short tandem repeat (STR) loci consistent with the AmpFℓSTR® Identifiler® Plus (Applied Biosystems®, Foster City, CA) and GlobalFiler™ (Applied Biosystems®, Foster City, CA) amplification kits were completed to explore the impact additional polymorphic loci have on estimating the NOC. The NOC for each profile was determined using the maximum allele count (MAC) method. An exponential or logistic regression of observed frequencies of dropout (Fr(D)) was found to be an appropriate characterization of allelic dropout rates. In general, the peak height based methods overestimated dropout at higher target levels and underestimated it at lower target amounts. The underestimation suggests that other factors beyond detection and polymerase chain reaction (PCR) variation contribute to dropout. Across all loci, the Fr(D) increased as target amount decreased and as molecular weight increased. Estimating the actual NOC using MAC was found to be unreliable for mixtures with greater than three contributors or with one or more minor contributors present at low levels. While a high level of dropout did not affect correctly identifying two-person mixtures, it greatly increased the number of misidentifications with three or more contributors. The number of misidentifications was reduced for mixtures when 21 STR loci plus amelogenin were used to evaluate the NOC. These higher accuracies were frequently attributable to the highly polymorphic locus SE33. The presence or absence of a major contributor did not appear to substantially affect the results. Forensic laboratories using MAC to determine the NOC of mixed samples should be aware of the tendency to underestimate the NOC using this method. It is also important to understand the impact that allelic dropout has on correctly estimating the NOC. The probability that allele dropout may have occurred in a sample should be considered when evaluating the NOC that explains the evidentiary profile