Inside the Cell, by Erin E. Murphy (Forensic DNA)

[Any of N]

Inside the Cell: The Dark Side of Forensic DNA / Erin E. Murphy - October 2015 - Nation Books [Link]

Inside the Cell, written by law professor Erin Murphy (NYU School of Law), discusses the civil liberties issues involved with forensic DNA testing. Although it functions as a kind of legal brief, you can ignore the politics if you so choose and focus on the wide-ranging information relevant to the EAR-ONS case. Murphy addresses so many of our endless questions about DNA that her book must surly qualify as must reading.

If you're not a science geek there's good news. The book is targeted at the educated layperson and represents a reasonably easy read. Murphy's undergrad was in Comparative Literature and not, say, Biology. On the other hand, the 400 pages are so densely packed with material that you'll be hard pressed to keep the details straight. I'll pass on trying to write a proper review and instead provide a quick overview of the key content areas relevant to our needs. Here is my arbitrary classification scheme.

Basics of DNA science

• DNA Typing
• The disputed odds of finding a hit
• Familial testing with DNA Typing
• Other types of forensic DNA tests: facial reconstruction, alternative familial tests

Lab Issues and Quality Control

• Touch DNA and transfer
• Contamination: samples that are mixtures of multiple people
• Degraded samples
• Labs not routinely reporting error rates
• Scandals at large labs

DNA Collection Protocols and Database Use

• CODIS search software
• Database Index Systems: National (NDIS), State (SDIS), Local (LDIS)
• Rules and standards for submitting and searching
• Rouge databases
  

The Laws Regarding DNA Testing

• Privacy and safeguards against fishing expeditions, or lack thereof
  
• DNA collection
• Use of private databases by law enforcement: 23andMe, Ancestry.com

That is a very incomplete list but it should give you an idea. My goal is to occasionally add to this thread with useful details. I've already excerpted from the book in multiple threads, but it will be useful to have it all in one place. As a final mention, the book is full of citations from journal papers, news accounts, and legal documents. Much like the main text, I've often found them to be useful and relevant to our discussions.

[debris]

I just downloaded the sample. Thanks for sharing !

- Deb

[Deleted]

Excellent. Amazon has used and new copies from $8-$13.00.

[Any of N]

Deb and Suburbs, you'll have to let me know what you think of the book.

Following through with my goal of summarizing key information taken from the book, here is a rundown on DNA typing.

DNA Typing / DNA Profiling:

• This is the type of forensic DNA testing we typically hear about.
• Designed to identify people. Like fingerprinting, but using genetics.
• It samples 13 different locations on the genome. The Latin words "locus" (singular) and "loci" (plural) are used.
• They are not considered to be "genes" because they are not thought to regulate bodily functions. They are part of the so-called "Junk DNA" of the genome.
• At each locus, there are two fragments, one inherited from each parent.
• Measurements are taken on each of the 26 fragments. They consist of "repeat lengths" of bases (allele patterns).
• A person's signature or genetic profile consists of the results of all 26 measurements.
• When DNA is compared across samples, an exact match consists of a match at all 26 allele patterns. Or on all 26 "alleles" for short.
• Note: The FBI is supposed to have expanded the number of loci used to 20 in January 2017 (last month).

CODIS and NDIS Database Searches

• CODIS is the "Combined DNA Index System" and is the software used to conduct DNA searches.
• The NDIS is the "National DNA Index System" kept by the FBI. It's the actual DNA database, not CODIS.
• The NDIS keeps records of DNA Typing profiles (the 26 measurements). No other types of DNA results are stored there.
• "NDIS labs must be affiliated with a criminal justice agency, be accredited, comply with quality assurance requirements issued by the FBI, and undergo periodic auditing for compliance."
• Profiles are indexed as coming from crime scenes, convicted offenders, arrestees, missing persons, unidentified human remains, and relatives of missing persons. There's also a miscellaneous legal index for any other profiles that don't fit the category and one for immigrant detainees.
• No identifying information is stored about the criminal case or the sample donor. That is kept at the state level.

Familial Searches Using DNA Typing

• A "CODIS search" means that we're talking about DNA typing.
• A familial search conducted with DNA Typing is simply a search for partial matches, i.e. less than an exact match at all 26 allele patterns.
• On average, full siblings will match on 17 or 18 alleles. That's according to one study. The range is 12-22.
• Large databases like the NDIS will yield a high number of partial matches for completely unrelated persons. (They can be classified as false positives.) Thus actual familial connections, parent-child and sibling-sibling, can easily get lost among the false positives.

Other genetic systems are much better for finding familial connections, such as those used by Ancestry.com and 23andMe. So when we hear about a familial DNA search, it's very fair to ask, "What type of familial DNA search?"

[debris]

I'll try to read some next weekend and get back to you.

[Any of N]

Four of the early paragraphs in Chapter 12, Genetic Informants: Familial Searches, are quite helpful. This is in regard to DNA Typing technology. I try to make these points all the time in our forum discussions. Here we will let the law professor make them for us.

First, the basics regarding CODIS familial searches. The degree of genetic overlap between parent-sibling and sibling-sibling relationships is not fixed. It's partially random and it will vary from one pair of relatives to another. (Emphasis added by me.)

To understand what is at stake, it is at first important to have a clear idea of how familial searching works. The basic premise is that persons who are biologically related to one another are more likely than unrelated persons to have similar genetic profiles. For instance, in a standard thirteen-locus DNA profile of twenty-six alleles, a child and a parent will match at a minimum of thirteen alleles. That is because the child gets half its DNA from its mother, and half from its father.

It follows that full siblings are likely to resemble one another genetically as well. Two full siblings might share no alleles in common—the unpredictability of inheritance means there is a 25 percent chance at each locus that one sibling inherits half of each parent’s alleles, while the other inherits the other half. But there is also a 25 percent chance that they share both alleles. On average, according to one estimate, full siblings have between 17 to 18 alleles in common, as compared to the 8 to 9 alleles unrelated persons are likely to share. The probability of overlap turns on several factors—most pertinently on common inheritance, but also on the likelihood that the parents themselves shared a particular allele and on the commonness of their alleles in the population at large.

Partial matches will occur among completely unrelated people. Results of this type qualify as "false positives."

Still, studies show that the larger the database, the more difficult it is to conduct an effective familial search. In one set of studies involving simulation searches of three state databases of varying sizes, the likelihood of finding a relative in the database dropped precipitously as the database grew. Whereas a parent or child in the database was listed within in the top ten candidate leads 98 percent of the time in a 19,300-person database, that likelihood fell to 72 percent of the time in a 356,000-person database, and to less than half of the time (48%) in a 1.8 million–person database.13 The estimate for the 10 million–person national database was that only 29 percent of the time would a parent-child relative be returned in the top ten leads. The numbers were significantly lower across the board for other relatives, such as full siblings.

Looser search criteria will lead to more instances of partial matches. Tighter criteria will lead to fewer matches.

For all of these reasons, the effectiveness of a familial search depends entirely on the specific parameters of that search. And there are tradeoffs. Narrow search parameters increase the likelihood that a lead found via such a search is, in fact, a relative of the perpetrator. But narrow parameters also increase the likelihood of missing out on that relative altogether because the criteria were too strict. Conversely, more generous parameters raise the probability that any relative in the database will be captured by the search, but this person may be buried among a long list of other entries that turn out to be false. In short, if the search standard is too strict, it may miss an entry that is related to the source; but if it is too loose, it will sweep in a larger number of false leads that make it more difficult to find the true link.

The big takeaways regarding familial searching via CODIS:

• Actual familial relationships can easily get lost in the noise.
• Close relatives are not guaranteed to have a close genetic match.
• Don't expect to find familial relationships that go beyond parent-child or sibling-sibling.

We have to look at other types of genetic tests for better familial searching.

[tim]

Any of N, you say that other genetic systems are much better for finding familial connections than CODIS. What types of familial DNA searches are there other than the (inferior) system that CODIS uses? For instance, what genetic tests do ancestry.com and 23andme.com use to find familial connections that is different from the CODIS method?

[Any of N]

tim : An answer to that can be made at various levels of abstraction. A geneticist (which I'm not) could take you so deep in the hole you'd regret ever asking. If you're like me, you just want a quick rundown of the key concepts. So here is how I understand it.

23andMe and Ancestry.com (and similar companies) provide a suite of tests for familial searches that include autosomal testing, Y-STR testing, and mitochondrial testing or 𐌼𐍄DNA. Autosomal chromosomes are inherited from both parents, Y-STR runs only through the male line, and 𐌼𐍄DNA runs only through the female line. It's all about the nature of the odds and the way the information is inherited.

Posters please feel free to correct me on this, but these three types of DNA testing encompass much more information than the 13 loci covered with CODIS (or even it's recent increase to 20 loci). So there will be less uncertainty in the results when looking for familial relationships. That seems to be the essential detail.

This blog seems to have a lot of good information regarding the commercial services.

dna-explained.com/which-dna-test-is-best/

[NaptimeNancy]

This explains so many questions I have had. Thank you for posting this! Not too long ago some friends of mine that are full blood sisters showed both their Ancestry DNA results and I thought it was so odd that their percentages within different origins were off by like 3-4%. I was sitting there side eyeing the results like, hmmmm you sure y'all have the same daddy? Haha. So interesting!

[Any of N]

Placing another quotation in this old thread. The benefit of adding Y-STR results to DNA Typing is something that didn't register with me initially but it eventually came up in discussion. It greatly improves the accuracy of familial searches, but only if it was also collected on every man who was DNA typed. The old question was if the State of California collected both.

Another way to narrow down a list of possible leads is to conduct additional genetic tests, such as to use certain signatures from the male chromosome that descend patrilineally (Y-STR typing). For instance, Colorado undertook a project to run 1,580 forensic profiles against its 87,000 known offenders database, generating roughly 137 million comparisons. Of those, thirty-two possible leads were produced: thirteen female and nineteen male. The male candidates were then further investigated using Y-STR typing, and six matched to Y-chromosome identifiers, suggesting the perpetrator was likely somewhere in their family line. Those six, along with the females, were then slated for further investigation.

Although effective, adding Y-STR typing presents its own complications. First, most databases do not currently include Y-STR data, and so physical samples must be retrieved and retested. Second, if there are a number of candidate leads, then multiple samples may require such re-testing, which adds both time and expense to the investigative process. And lastly, reliance on Y-STR typing limits the utility of familial search methods to identifying potential leads in the male population only; a perpetrator who has only a female relative in the database may still go undetected.

Erin Murphy, Inside the Cell: The Dark Side of Forensic DNA, Chapter 11, Sneak Sampling, Dragnet Searches, and Rogue Databases (2015).

The superior approach to familial searching is through autosomal DNA testing, as was done to locate Joe DeAngelo along with genealogy tracking.