Comments on: Ancestry painting: true but trivial, or interesting but inaccurate

By: chipoltespice

chipoltespice — Sun, 22 Apr 2012 01:35:31 +0000

If Asian is related to Native American, segments are missed and those segments are painted the predominate ancestry. Sending to Doug McDonald will help. Due to recombination especially with a non-Asian parent, inheriting smaller Asian is not abnormal.

By: Razib Khan

Razib Khan — Thu, 19 Apr 2012 13:45:12 +0000

I was under the impression though that IE9 was a step forward in cross platform app friendliness

right. but the bar was very low. like american cars in the 1980s IE now has its past history to deal with. though whenever a new version comes out i take it for a test drive. nothing i’ve seen would make me want to switch to IE.

By: April Brown

April Brown — Thu, 19 Apr 2012 13:41:21 +0000

Re: the browser stuff… well that’s all just colossally depressing. I used to work with the IE team, and was up on all this, and here is yet another example of motherhood making me dumb and ignorant.

Back when Chrome came out as a public beta there were horrific security issues (by design), with Google using the browser to send a lot of data back to headquarters, probably in the name of usability research. I think that’s since been addressed, but clearly I’ve been out of the loop so long that I’m not much of an authority.

I was under the impression though that IE9 was a step forward in cross platform app friendliness and not being such a resource hog. I can understand not wanting to be compatible with IE7 or IE8.

By: Eurologist

Eurologist — Thu, 19 Apr 2012 10:27:43 +0000

And yeah, we can’t deal with IE and its nuances, we just went for the most open solution, which meant Chrome (and FF where possible… all things considered, this is a side project and no one’s day job yet).

Smart decision (I was a principal in several start-ups in a prior life – I know exactly how resource-intensive this is, including QC on numerous OSs).

Kind of amusing that Interpretome isn’t compatible with Internet Explorer. Bit of a silly limitation.

It only takes a few minutes of your time to install a decent browser – but it may take a full-time position or more at a (presumably) underfunded organization to guarantee multiple browser compatibility.

By: ackbark

ackbark — Thu, 19 Apr 2012 00:15:06 +0000

Interpretome certainly shows more variety than the 23andMe ancestry painting, but it classes the bulk of me as Tuscan rather than CEU which is more what I’d expect as all other information I have says Northwest European.

By: Konrad Karczewski

Konrad Karczewski — Wed, 18 Apr 2012 23:05:27 +0000

Interpretome inventor weighing in here. Thanks for the shoutout! Glad to see the experimental Hapmap 3 is working for someone, it’s a bit odd because there are some pretty similar populations in there (and we’re concerned the overlap may be producing some weird results). We’re tinkering with some other reference populations, so that might help a bit. As a side note, cranking up the “Number of Samples” parameter can only help, it just takes longer because your browser is doing some more haplotype sampling.

By: Razib Khan

Razib Khan — Wed, 18 Apr 2012 14:11:56 +0000

#8, for client side development IE is a shitty browser. not amusing at all, cross-browser design wastes an enormous amount of time. i’d rather have this sort of niche application focus on extending the feature set on one browser than crippling itself to work on multiple browsers. chrome’s original stated purpose was to keep the major browsers ‘honest,’ but from what i recall it surpassed firefox recently, and is catching up to IE.

By: April Brown

April Brown — Wed, 18 Apr 2012 14:10:29 +0000

Kind of amusing that Interpretome isn’t compatible with Internet Explorer. Bit of a silly limitation.

By: AG

AG — Wed, 18 Apr 2012 13:20:01 +0000

Meiosis (recombination), selective sequencing by 23andme,

By: Lank

Lank — Wed, 18 Apr 2012 12:45:56 +0000

The lack of reference populations isn’t the only problem. The threshold for what’s classified as “European” seems to be significantly lower than for the two other categories. 23andMe’s Ancestry Painting is very bad at picking up ancient admixture events. This is less obvious for samples with, e.g., an “Asian” affinity as high as yourself. But when that’s diluted with additional European, much of what was previously classified as “Asian” becomes “European”.

East Africans have similar issues; the amount of African in the Ancestry Painting isn’t proportional to the SSA (Yoruba) affinity, even though the African category is based on the Yoruba.

By: Eurologist

Eurologist — Wed, 18 Apr 2012 10:20:59 +0000

“While unlikely the 8% Asian would be one possibility with a 40% Asian parent. ”

There are (possibly many orders of magnitude, depending on how you count) differences between “possible” and “unlikely” – or extremely unlikely. Ancient autosomal DNA isn’t going away quite that quickly, with increasing (or full) resolution.

By: Razib Khan

Razib Khan — Wed, 18 Apr 2012 07:13:31 +0000

#3, i’m 90% sure this is just a problem of lack of reference populations in the painter. mostly resolved when they were added. so it would be easy to found out about native ancestry in the case you are talking about (or lack thereof genomically).

By: Dwight E. Howell

Dwight E. Howell — Wed, 18 Apr 2012 05:53:01 +0000

While unlikely the 8% Asian would be one possibility with a 40% Asian parent.

That’s why I’m sort of neutral to some of the data I know about my family tree. I know I have Cherokee ancestors through a female. I also know that was several generations ago so what are the odds of me or my younger relatives having any native American genes? Low but distinctly possible.

The same sort of situation is faced by a local family that knows one female ancestor came off a plantation after the civil war. Does anyone in the family still have any of her genes? Maybe but they’ve been thoroughly diluted by now.

Of course this also show’s why trying to build a complete family tree based on present genetics is not to be taken seriously once you go back past a certain point. You are clearly going to have ancestors whose genes never made it to you.

This pretty much explains what I think happened to Hsn. Europe seems to be one of those places into which populations constantly flow and the local genes are constantly getting replaced or diluted until little or nothing is left. I don’t see much reason to think the people doing the replacing are in any meaningful sense superior to those being replaced as long as they keep coming in wave after wave which seems to be what actually happens and is still happening.

Time to go to bed.

By: Razib Khan

Razib Khan — Wed, 18 Apr 2012 04:57:19 +0000

#1, select hapmap3, smoothing down to 10, chi-square down to 0.10. left other stuff alone. using a broader set of reference populations was the essential thing though.

By: Doktor Grisha

Doktor Grisha — Wed, 18 Apr 2012 04:54:03 +0000

So, you did not use the default settings in Interpretome. Well, then, Razib, publish your parameters!