The field of forensic science has seen tremendous growth in the last few decades. CSI and other TV series have helped to popularize the use of DNA samples to catch a suspect. DNA evidence, however, is not as cut and dry as modern television would have us believe.
Rutgers-Camden professor Desmond Lun, in partnership with a team from Boston University and MIT, is trying to solve one of the biggest issues encountered with DNA evidence. The group recently received a $206,946 grant from the National Institute of Justice to develop a software program that will use statistical analysis to determine the usefulness of a particular DNA sample to a criminal investigation.
DNA evidence is often simplified in modern culture, but the forensic science community has been embroiled in a debate for the past decade about how to analyze samples that may be contaminated by DNA from people other than the subject.
In a perfect situation, Lun says that crime scene investigators collect a DNA sample from the scene, analyze its genotype and narrow the list of suspects down to a small class of individuals.
Unfortunately, this is rarely the case, Lun said. Most frequently, investigators “actually get a sample that is corrupted by a lot of contributors.”
Until now, there have been “no good answers,” said Lun. That is why he believes their grant proposal was successful. It is an “unusual theme” that will give investigators the ability to quantitatively analyze potentially contaminated DNA evidence.
Contaminated DNA samples are referred to as “noisy” samples. Imagine having a conversation in a noisy room, he explained, but you and the person you are talking to are on opposite sides of the room. All the background conversations make it difficult to hear what the other person is saying because you also get snippets of every conversation in the room.
In the CSI world, “other conversations are DNA from unidentified people” who have been at the scene but are unrelated to the crime. The signal that investigators want, however, is DNA from the suspect. When a crime scene has been heavily traveled, the suspect’s DNA can be mixed up with samples from every person who has been at that location.
Noisy samples are problematic because they “involve a bit of guesswork” which means that “conclusions are substantially influenced.”
“The majority of samples have some level of corruption,” so while they may contain useful information, it is difficult to narrow down what is and what is not useful. Corrupted evidence is subjective and can improperly influence the outcome of a trial, if it is even allowed to be presented. This is where Lun’s proposed statistical analysis software becomes important.
Lun’s background includes engineering and computational expertise. His partners at Boston University and MIT are chemistry and forensic science experts with “strong backgrounds in analyzing difficult signals,” Lun said.
Right now, they are working on creating the right calculations. The team will then build a mathematical model, assigning probabilities that can calculate different random events and compare the random ways that a particular DNA sample could present itself in an event. When this information is calculated, the software will issue a probability that a particular sample came from a subject of interest. The method will allow investigators to determine “how confident we are that the DNA belongs to a particular subject.”
Lun expects that a prototype of the software will be available within the next two years.
“We’re not aiming for a complete, finished and polished program” right away, Lun said. The plan is to train crime lab technicians how to use the free program. If it is successful and appealing, they will develop a more polished version that would require minimal training.