This came up in a more or less public chat I recently had with Henry Zenker on the R-P312 Yahoo group (June 21). I was referring to the notion that matching off-modal STR values, in two otherwise not very closely matched haplotypes, could be perceived as analogous to the way notches on a key have to match the tumblers in a lock. He replied, "Thanks for that STR / lock tumbler analogy. It's a strong tool to use to refine my thinking regarding STRs. For the toolkit for sure."
The concept arose today in another conversation with Henry; and I mentioned that I had gotten that lock-and-key idea from a dendrochronologist I knew in the 1980s, Herman J. "Jack" Heikkenen. I did a little research and found a nice 2002 article about his work in the journal of the Colonial Williamsburg Foundation. It's available online, so I guess it's permissible to quote a passage; anyway, here is the source, after which I'll paste in a couple of key paragraphs (pun intended):http://www.history.org/foundation/journal/spring02/dendro.cfm
Jack Heikkenen’s great addition to dendrochronology is the “key year” concept of tree growth. “The whole system comes down to this,” Heikkenen says. “Each year’s growth ring, compared to last year’s, can be only one of three things: greater, which is recorded as a plus; equal, which is a zero; or less, which is a minus. Because you get so few zeroes, the result is a plus or minus pattern for the annual rings in any given tree.
“Then, if you look at a sample of trees from a certain locale, you may notice that twelve out of fifteen will all show a plus or a minus for a particular year. That will be a ‘key year.’ The pattern of key years can be matched up with the pattern in the wood of a house whose date is in question.” A computer calculates the fit and probability of error for the date the wood was cut.
All together, Heikkenen’s system ranges from boring hollow-bit cores out of house timbers, to polishing the cores, to measuring each ring in the cores under a microscope, to pushing a button that declares whether a ring is a plus, minus, or a zero, and finally to dumping all this into a computer that compares the strings of pluses and minuses with data on local or regional trees. Heikkenen has patented the system, and it’s changing the way we look at our colonial past.
If I have any contribution to make here, it's just to call attention to the technique; and to the fact that the colorized results in our large projects function in substantially the same way as tree rings. Modal values are widely shared, and while those who share them are alike, that is of little interest. We might consider the haplogroup's modal as the grooves on a blank key that enable it, for example, to slide into a Master lock, as opposed to a Schlage lock. Important, but we already know it. The useful information is the pattern of red (plus) and blue (minus) results. Within reason, the more markers tested, the more shared off-modals one will have with another person who is related at the recent subclade level. (Say, within the past two or three millennia.) For the most part we will match in color (above or below modal) and intensity (about the same number of steps
up or down, at a significant locus), at many loci. And there will be some consistency in our inconsistencies: when we mismatch, it will prove to be at fast-mutating markers. The simple visual comparison is very much more powerful and predictive than counting the "genetic distance."
As an aside, I'd like to mention that upgrading from 37 to 67 markers brought me very few new off-modal values. But upgrading from 68 to 111 brought many more; and for the most part they appear to make finer distinctions between my closer, or more distant, genetic "relatives." Inability to compare with a large data set is still a drawback, but as more people test out to 111 the advantages are increasing nicely.
Anyway, this is proposed as a quick and dirty mechanism for determining what SNP to test next. In any big project, set the results to "colorized;" find the pattern of blues and pinks that best matches your own (but mismatches most other people's); and test whatever terminal positive SNP that guy has.