Cold Case Hits Use Vastly Exaggerated DNA “Match” Statistics; Upheld by California Supreme Court

by Matt Clarke

A recent California murder trial has highlighted serious problems in the probability statistics used to determine the odds of DNA matches in cases that involve DNA database searches. However, the California Supreme Court upheld the admission of evidence regarding the process used to obtain those odds in a separate case dealing with the same issue.

In December 1972, Dianna Sylvester was raped and murdered in her San Francisco apartment. The case went unsolved for decades. Meanwhile, in 1977, John Puckett committed two rapes and a sexual assault in the Bay Area. He was convicted and sentenced to prison. Other than a 1988 misdemeanor battery charge, his record has been clean since his release from prison in 1985. The obese, 70-year-old, wheelchair-bound Puckett recently had triple bypass surgery.

In 2004, California gave San Francisco funding for DNA testing in cold case murders. A swab containing sperm taken from Sylvester’s mouth was retrieved and tested. The usual number of genetic markers used in DNA testing is 13; such markers contain less than a millionth of the information in a human DNA molecule. In Sylvester’s case, only 5 ½ genetic markers were usable, and they were mixed with markers from another person, presumably Sylvester herself. That was too little information to search California’s 338,000-person DNA database. However, an analyst added another 2 ½ markers to obtain the minimum number necessary to scan the database, by extrapolating from markers that were so faint they were inconclusive. The database comparison using the extra extrapolated markers turned up a DNA hit on Puckett.

At his trial in January 2008, there was no direct evidence to tie Puckett to the 1972 crime. The DNA hit, his criminal record and the fact that Puckett was in the Bay Area when Sylvester was killed was the only evidence linking him to the murder. None of the 28 sets of fingerprints found at the crime scene matched Puckett.

One of the prosecution’s DNA experts testified that the chances of a random match to the DNA recovered from Sylvester was 1.1 million to one. Another gave his own “likelihood ratio” as 1 in 152 billion. The jury convicted Puckett, but only after sending a note to the court asking how he had become a suspect – a question the judge refused to answer.

At a pretrial hearing, Bicka Barlow, a DNA specialist for the San Francisco public defender’s office who has a master’s degree in genetics, had testified that the match probability statistics were wrong. The problem was that the 1 in 1.1 million statistic applied to a random match to a single individual. That doesn’t hold true when you began comparing DNA profiles with huge databases of DNA samples. Then, according to two panels of scientific experts convened by the FBI and the National Research Council, the chances of a random individual match should be multiplied by the number of samples in the database. That would change the odds of an innocent person receiving a DNA hit in Puckett’s case from 1 in 1.1 million to approximately one in three. Yes, one in three.

But the judge in Puckett’s trial never let the jury hear about this statistical disparity. He ruled it was inadmissible. He also ruled that the jury could not hear about the person the police originally suspected in Sylvester’s murder before the DNA hit on Puckett. Police had arrested Robert Baker, a street artist who had escaped from a mental institution, for a rape that occurred four blocks from Sylvester’s apartment two weeks before her murder.
He was also identified as having harassed and followed a young girl and woman to their home a few houses down from Sylvester’s apartment four days after the murder. A blood-spotted parking ticket was found in Baker’s van when he was arrested. However, Baker, who died in 1978, was never charged with the murder; the parking ticket was lost and usable DNA from Baker is unavailable.

The problems with cold case hits using DNA databases are well known to experts in the field. “It’s only a matter of time until someone is wrongfully convicted because of this,” said Stanford mathematician Keith Devlin.

The concern is not merely theoretical. In 2001, a six-marker DNA profile of Raymond Easton resulted in a cold hit in Britain’s extensive DNA database. Easton was arrested and charged with robbery; it was then learned that he lived 170 miles from the robbery and had a rock-solid alibi. He was exonerated after further testing of additional DNA markers.
The chances of a random DNA hit in Easton’s case was 1 in 37 million. But it was only 1 in 57 when the size of the DNA database was taken into account.

The jurors in Puckett’s case found the DNA probability statistics convincing. They went with the 1 in 1.1 million odds as the most “credible” and conservative number, and said that was pivotal in reaching a unanimous verdict. Would the outcome have been different had they known about the one in three chance of an erroneous match?

“Of course it would have changed things,” said juror Joe Deluca. “It would have changed a lot of things.”

Meanwhile, Puckett sits in prison with a life sentence for Sylvester’s murder. And there may be many more questionable cold case DNA hits in the works. There are about 6 million DNA profiles in various databases throughout the country. They have already generated 50,000 cold case matches, most of which have not yet gone to court. And when they do, at least in California, those defendants will face tough odds.

On June 16, 2008, the Supreme Court of California considered the use of DNA in a 26-year-old murder case and the method used to calculate the odds of an incorrect match.
The Court noted that the “product rule” used to determine the odds of a DNA match “has gained general acceptance in the relevant scientific community and therefore meets the Kelly [People v. Kelly, 17 Cal.3d 24, 549 P.2d 1240 (Cal. 1976)] standard for admissibility.”

In this case, the DNA match occurred in a database containing around 184,000 DNA samples. The Court held that the use of the product rule to determine the odds of a DNA match in a cold case was a “a question of relevance, not scientific acceptance,” since the product rule had already passed the Kelly test and had been established as admissible evidence. The Supreme Court further found that “although the product rule is not the only available method of statistical analysis in a cold hit case, it is relevant and thus admissible.”

While the government had conceded in a previous case “that in a cold hit case, the product rule derived number no longer accurately represents the probability of finding a matching profile by chance,” the Court differentiated between the odds of obtaining a DNA match from a database search and the rarity of the specific match in society as a whole. That is, the rarity odds were consistent whether the match occurred through a database search or not, and thus were relevant and properly admitted as evidence.

“The fact that the match ultimately came about by means of a database search does not deprive the rarity statistic of all relevance,” the Court ruled. “It remains relevant for the jury to learn how rare this particular DNA profile is within the relevant populations and hence how likely it is that someone other than defendant was the source of the crime scene evidence.” See: People v. Nelson, 43 Cal.4th 1242, 185 P.3d 49 (Cal. 2008), cert. denied.

The fact remains, however, that there is a very high statistical probability of an incorrect match when using large DNA databases for comparison purposes – which is how cold case DNA matches are obtained – and that most jurors will never hear about that very relevant discrepancy.

Source: Los Angeles Times

Subscribe today

Already a subscriber? Login

Related legal case