Establishing Exclusion Criteria and the Significance of Inclusion for Complex Low-Template DNA Mixtures

DNA mixtures may be comprised of any number of contributors and combined in any proportion. Current interpretation schemes require laboratory personnel to validate and determine analysis parameters for each method employed. In many cases, commonly employed DNA interpretation protocols do not allow for accurate comparison between standard and evidence. There are two general approaches to interpreting DNA profiles. One is the random man not excluded method, which is a binary scheme that assumes alleles are either absent or present; the other is the likelihood ratio method, which assesses the DNA evidence probabilistically to compare a prosecution hypothesis and a defense hypothesis. Both methods require assumptions to be made prior to analysis and may not be suitable for multi-contributor, low-level samples. Further, the signal obtained from biological evidence may be extremely complex and the signal obtained from the person-of-interest may not be distinguishable from baseline noise, PCR/instrument artifact and multiple 'other' contributors. In addition to likelihood ratio tests that compare two hypotheses to determine the number of contributors, we intend to examine the likelihood of the evidence directly - assuming multiple contributors exist. Specifically, we propose to introduce matched filter techniques from the digital communication field as a tool for generating likelihoods for hypothesis testing. A matched filter is obtained by correlating a template signal with an unknown signal to identify the presence of the template signal. It is the optimal linear filter for maximizing the signal to noise ratio in the presence of additive noise, and is commonly used in Code-Division Multiple-Access (CDMA) modulation, which is the basis of multiple-user detection in 3G radio communication systems. In the realm of DNA signal detection, the DNA profile of an individual of interest is used as the template signal, and by correlation, one may compute its distance to the DNA mixture under investigation. Such correlations between the crime profile and reference profiles do not require the number of contributors to be known a priori, since for each matched filter, contributors other than the one of interest are counted as noise (or interference). Comparison among the output of bank of matched filters then ranks the likelihood of each template DNA profile. Unlike existing mixture interpretation schemes, the matched filter approach does not identify or attempt to infer the genetic profiles directly. Instead, by systematic comparisons of the crime scene profile to those from individuals of interest, it is possible to include or exclude individuals as potential contributors while providing statistical weight to that conclusion. ca/ncf