Abstract

Statistical and Probabilistic Issues in DNA Forensics - Current Paradigms

Ranajit Chakraborty, PhD

Department of Environmental Health

University of Cincinnati College of Medicine, Cincinnati, OH

DNA Forensics deals with use of recombinant DNA technology for forensic investigation. On a routine basis, DNA forensics is practiced for human identification, kinship analysis, missing person identification through comparisons of DNA profiles of biological samples collected from crime scene, human remains, etc. with those of one or more known persons. Interpretation of matched or concordant profiles (at times even partial ones) is based on probabilistic principles and statistical modeling of data. Since DNA evidence is generally composed of low probability events, statistics for DNA forensics confront issues of sparse data, and estimation of multidimensional small probabilities. This presentation starts with a brief history of developments in DNA forensics, and with examples of generic classes of DVA evidence data, discusses three approaches of statistical assessment of DNA evidence (Frequentist, Likelihood, and Bayesian), showing that these logic are built on each other, and each of them answers somewhat different, but related questions. Issues of database size and population subdivision effects in estimating relevant probabilities are also addressed. Currently practiced paradigm is also presented with discussion on comparison of observed and expected matches (full as well as partial) in database searches. Finally, while mitochondrial DNA and Y-chromosome haplotypes are argued as important supplement of autosomal DNA loci, differences of statistical approaches for such lineage markers from those for autosomal markers are exemplified. This presentation sets the stage of three other types of applications of probability theory in DNA forensics that are presented by three experts in the field in this symposium.