10 Forensic Magazine | www.forensicmag.com MARCH 2018
Hidden DNA Evidence:
Exonerating the Innocent
Mark W. Perlin, Ph.D., M.D.
In 1989, northwest Indiana was plagued by bump-and-rob road crimes of escalating violence. In the darkness of a cold December night, a woman’s car was rear ended on highway I-65. Upon exiting her car, she was dragged into another vehicle, then stripped and raped by five strangers. The men left
her in her car, draped by green coveralls. The same night, coworkers Darryl Pinkins, Roosevelt Glenn
and William Durden had engine trouble along that highway. They parked their car on the roadside,
and went to get help and motor oil. On their return, they found shattered side windows and their work
Traced to their employer by the crime scene coveralls, Pinkins, Glenn and Durden, along with two
other coworkers, were arrested for the I-65 bump-and-rape. RFLP testing of semen DNA left on the
victim’s jacket and sweater excluded the defendants. But nonspecific serology testing, along with faulty
hair evidence and tainted eye witness identification, led to Pinkins’ and Glenn’s wrongful convictions.
Pinkins was found guilty of all charges in May 1991, and sentenced to 65 years in prison. Glenn’s 1992
jury deadlocked, but on retrial he was convicted of rape in 1993, and sentenced to 36 years. Despite
their incarceration, the bump-and-rob and rape crimes continued unabated. The men’s exoneration by
science would not happen soon.
In the early 1990s, the Human Genome Project was in full gear. A key genetic marker was the short
tandem repeat (STR). At an STR genetic location (locus) on a chromosome, naturally occurring varia-
tion in the number of repeated words (allele) could track inherited DNA or distinguish different people.
This allelic repeat variation was easily
measured on a DNA sequencer—more
repeats gave longer sequences, and
more allele DNA molecules fluoresced more brightly. STR data have
a “stutter” artifact, where an allele
produces shorter shadow sequences.
The stutter problem was solved mathematically by removing these DNA
shadows. This genotyping automation
was immediately applicable to genetic
diagnosis, gene discovery and human identification. Computers could
interpret STR data faster and more
accurately than human review. In
fact, automated DNA review software
eliminated Great Britain’s national
DNA database backlog of 350,000
offenders. Each cheek swab contained
abundant DNA from one person. But
Figure 1: Match statistics between evidence genotypes (rows) and reference individuals (columns). The entries give the base ten logarithm of the match statistic.
For example, + 6 log units would represent an inclusionary million (six zeros after
the 1) statistic, whereas – 6 log units would represent an exclusionary one in a