University College London's David Balding argues in the Proceedings of the National Academy of Sciences that the statistical methods used to evaluate noisy, mixed-source, low-template DNA profiles have "serious shortcomings."
Such profiles, he notes, have been used in court cases in the UK and in Italy, where the methods become the center of controversy, such as in the trial of Amanda Knox, New Scientist notes.
Balding instead proposes a new method, called likeLTD, that calculates a likelihood score that certain people contributed to crime scene DNA samples. The calculation takes different drop-out and drop-in rates, different amounts of DNA from different contributors, degradation, and more into consideration. His method suggests it is unlikely that Knox contributed to the crime scene sample.
University of Oslo forensic geneticist Peter Gill tells New Scientist that that the software is based on theories he developed a decade ago, but that have been slow to enter the courts. "In order to gain court acceptance I argue that transparency of software is required and therefore the way forward is open source," Gill says. "Balding's software [which is open source] satisfies this criterion."