Forensic Statistical Tool (FST): A Probabilistic Genotyping Software Program for Human Identification

The forensic statistical tool (FST) is an in-house forensic software package that was developed by Dr. Adele Mitchell and validated by a team of forensic analysts at the Office of the Chief Medical Examiner (OCME) Laboratory in New York City.1 This software and the data generated from it has recently come under question as part of a larger issue to obtain full disclosure to independently assess its use for interpreting complex DNA mixtures and appropriate likelihood ratio (LR) statistics in the criminal justice system.2 The purpose in developing this software was to take complex DNA mixture data from criminal casework and assess the likelihood that a given particular individual was included or excluded in the sample data along with a certain number of unknown potential contributors to the sample. This software uses predictive modeling based on a variety of settings chosen by the analyst at the time of software analysis and has the ability to show a correlation (inclusion) or no correlation (exclusion) numerically for a report. The adjustable settings include number of contributors (2 or 3 sources based on number of alleles detected at loci across a profile) and degraded or not degraded (for quality of DNA) and thus, the LR value is conditioned on number of contributors, quality and quantity of DNA detected. For analysis, FST uses a semicontinuous model and does not take into account stutter, masking of alleles, or genetic sharing of alleles by relatives. A qualitative rating scale is used to give a verbal description of the strength of the numeric value in relation to other LRs generated by the FST software. Examples include an LR of 1.0 (no conclusions can be drawn about inclusion or exclusion to the mixture), 1 to 10 (limited support for inclusion), 10 to 100 (moderate support for inclusion), 100 to 1000 (strong support), and greater than 1000 (very strong support). Negative numerals give weight to exclusions with a scale of -1 (limited support for exclusion) to -1000 (very strong support for exclusion) for an individual not being represented in the mixture