Back to Publications
DNA Mapping the Crime Scene: Do Computers Dream of Electric Peaks?
Perlin, M.W. DNA mapping the crime scene: do computers dream of electric peaks? in the Proceedings of Promega's Twenty Third International Symposium on Human Identification. Nashville, TN, 2012.
Items of evidence, from one or more scenes, can fit together to tell a story of crimes and their victims. A hat, a gun, a cell phone or countertop - whatever was touched, handled or worn - can build an invaluable investigative map of criminal activity. In an age of "CSI effects", corroborative DNA may prove vital in persuading a jury.
But this investigative dream has become a crime laboratory's nightmare. Low-level touch DNA mixtures stream in through the door, along with investigator expectations of fast, informative results. People and robots churn through challenging evidence, cranking out complex EPG signals that defy human interpretation. Key DNA is often deemed "inconclusive", and many comparisons must be made to find a probative match.
Enter the computer. In accordance with SWGDAM DNA interpretation guidelines (1, par. 3.2.2), computers can unmix the mixtures, separating out the data into each contributor's (probabilistic) genotype. The stochastic uncertainty of low-level STR peaks can be reliably modeled through statistical computing (2). Probabilistic genotyping can thoroughly examine the data, finding highly informative DNA match statistics.
Comparing many mixture items with many suspects across multiple cases leads to a very large number of potential matches. But perhaps only 10% of these genotype comparisons will actually yield a probative result. A computer can map out crime scenes by comparing evidence genotypes with references to implicate suspects. With computer assistance, a forensic analyst can translate an investigator's dream into evidentiary reality.
In a recent criminal case, a gang was suspected of committing a series of convenience store robberies. As the violence escalated, resolving these crimes became increasingly important. Touch evidence items were collected from five shops, yielding a dozen DNA mixtures. The STR peak heights were low (many under 50 rfu), and the mixture complexity was high (most items having 3 or 4 contributors). Comparisons were needed with ten reference genotypes, half suspect and half victim.
TrueAllele® computer interpretation of the touch DNA was able to separate evidence data into probabilistic genotypes. Eight of the twelve mixture items matched a reference, with DNA statistics ranging from thousands to quintillions. Three suspects were identified, including the likely gang leader. Whereas human assessment of the 120 potential DNA matches (12 evidence vs. 10 references) was not feasible, the computer could rapidly pinpoint the 9 probative matches.
Over a hundred criminal cases have been solved in this way, both for prosecution and defense, most involving complex DNA mixtures. Computer separation of mixed data into contributor genotypes preserves the identification information of examined DNA evidence items. Crime labs are now introducing these interpretation systems into their workflow as just one more validated analytical instrument. Their forensic scientists can use computers to convert "inconclusive" findings into solid DNA matches that map out crime scene events.
- SWGDAM. Interpretation guidelines for autosomal STR typing by forensic DNA testing laboratories. 2010.
- Perlin MW, Legler MM, Spencer CE, Smith JL, Allan WP, Belrose JL, Duceman BW. Validating TrueAllele® DNA mixture interpretation. J Forensic Sci. 2011;56(6):1430- 47.