tag:blogger.com,1999:blog-3023805782808412230.post8016556507983385501..comments2024-03-09T15:46:44.638+01:00Comments on For what they were... we are: More North Iberian Epipaleolithic mtDNA (and first Epipaleolithic nuclear DNA)Majuhttp://www.blogger.com/profile/12369840391933337204noreply@blogger.comBlogger149125tag:blogger.com,1999:blog-3023805782808412230.post-32313666631953063422013-07-11T21:02:46.402+02:002013-07-11T21:02:46.402+02:00ADMIXTURE is based on STRUCTURE that has been arou...ADMIXTURE is based on STRUCTURE that has been around for about 10 years. Actually for small datasets my gut feeling is that STRUCTURE still outperform ADMIXTURE. These are powerful but the technology move forward. Today we have software like Chromopainter-Finestructure that is able to work with haplotypes rather than single SNP allle frequencies. IAnders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-19552321540282738802013-07-11T16:29:56.203+02:002013-07-11T16:29:56.203+02:00Oops, it's not Magdalenian but Epigravettian (...Oops, it's not Magdalenian but Epigravettian (hence related to Italy). Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-49910409477665824782013-07-11T16:28:04.267+02:002013-07-11T16:28:04.267+02:00OK, I take notice of your observations. They are p...OK, I take notice of your observations. They are potentially interesting but I rather prefer to wait for new aDNA sampling in SW Europe. For example, a rare open-air Magdalenian burial was <a href="http://cuevadelapileta.blogspot.com.es/2013/07/arqueologos-descubrieron-una-tumba-de.html" rel="nofollow">just discovered in France</a> and I'm hoping (no news yet) that they do test the DNA. Several sample points will be no doubt more informative than just one low quality composite. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-36237263769281743242013-07-11T13:31:29.689+02:002013-07-11T13:31:29.689+02:00The K12b calculator analysis by Dienekes using sub...The K12b calculator analysis by Dienekes using sub 1k number of SNPs is by the way "fluked". I have run my La Braña dataset vs the K12b calculator using 2.5k matching SNP's and got the following:<br /><br />North_European 77.34<br />Atlantic_Med 21.19<br />Southeast_Asian 1.47<br />Gedrosia 0<br />Siberian 0<br />Northwest_African 0<br />South_Asian 0<br />East_African 0<br />Southwest_Asian 0<br />East_Asian 0<br />Caucasus 0<br />Sub_Saharan 0<br /><br />As we can see the major "North-European" is far higher using 2.5k SNP than Dienekes sub-1k SNP dataset and as we can see from the minority components they have dissappered. This is the effect of small number of SNP as I have also tested with other genome calculators and the results are very variable for the smaller components shifting between East-Asia, African and Native American but very consistent having huge Northern European ancestral component. <br /><br />However in the 68k based analysis there is no doubt that there is really a African-like ancestral component.Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-14557540455909067812013-07-05T15:26:33.034+02:002013-07-05T15:26:33.034+02:00I do have practical experience with ADMIXTURE, whi...I do have practical experience with ADMIXTURE, which is also a long-time tested tool of which we can know from observation (and also "theoretical" readings) how it behaves in real life. Sample quality and size matters a lot and homozygous populations do tend to distort it so much that, for general analysis purposes, they need to be removed or minimized in weight (sample number). That is a fact. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-14390057584200615842013-07-05T08:29:36.014+02:002013-07-05T08:29:36.014+02:00I appreciate your skeptism, but I would even appre...I appreciate your skeptism, but I would even appreciate your skeptism more if you also had practical experience and not only theoretical experience with the tools I am using. Being sceptical just to be skeptical is not useful or helpful if not backed up by actual knowledge.<br /><br />Maju wrote:<br /><br />Which I'm being skeptical about. Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-54934634130113769372013-07-05T08:18:58.634+02:002013-07-05T08:18:58.634+02:00ADMIXTURE is about allele frequencies and you need...ADMIXTURE is about allele frequencies and you need some more samples to get right clustering and inferring the ancestral alleles closest to the truth. In Chromopainter-Finestructure its much less important to have large samples but it is helpful to have at least some to get the tree structure and heatmap boxes right. I have shown this numerous times in Chromopainter-Finestructure where it is very easy to separate Saamis and Finns from each other, but it is not so easy in ADMIXTURE with small Saami sample size.<br /><br />*Maju wrote:*<br /><br />"How can that be an advantage? It will totally distort the results"Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-36103635558322694112013-07-05T08:07:51.085+02:002013-07-05T08:07:51.085+02:00As the post titles says its about world variation,...As the post titles says its about world variation, I believe it was mentioned in the introduction too. So how you can misunderstand this I do not know and I do not know any other way how to infer African or Siberian admixture in Europeans without samples from these populations. I am interested in how the La Braña and European populations differ from each other vs the rest of the world, not about the internal relationship in Arabia or East-Asia, others can look at those questions.<br /><br />So yes again what I call the Middle-East component peaks among the Bedouins. I believe this have been shown in many research papers using the HGDP population panel.<br /><br />Maju wrote:<br /><br />"How can you be so sure that this component has anything to do with Bedouins and that it actually peaks among them if you did not use them in the run?"Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-83369170961436724532013-07-05T05:14:59.830+02:002013-07-05T05:14:59.830+02:00There are no Bedouins in the graphs you published,...There are no Bedouins in the graphs you published, and Sardinians hold the greatest apportion among your published samples, followed by Turks, Georgians, Italians, Basques, Spanish, Bulgarians...<br /><br />How can you be so sure that this component has anything to do with Bedouins and that it actually peaks among them if you did not use them in the run? Notice that components come and go depending on populations sampled, etc. and other components found in other analysis need not need to be directly extrapolable.<br /><br />"He had the advantage of having large Saami sample"...<br /><br />How can that be an advantage? It will totally distort the results. Exaggerating a bit for the sake of clarity, it's almost as if you introduce a large sample of clones, which are a component by default, dragging all others to position in relation to them. I guess you could even get several components out of large enough sample of true identical clones. <br /><br />Or to put another example, let's say I gather my "army" of almost 40 cousins and sample them, then compare with five others from here and two others from there. Don't you think that most of the results will be about slight differences between my relatives, rather than about differences between them and the unrelated minor samples? I am pretty sure that it will work that well. So a large Saami sample is almost for sure not any help but a burden. <br /><br />"so I couldnt replicate his analysis using ADMIXTURE based tools With so few samples"<br /><br />Your non-replication with a more correct number of samples actually seems to me the CORRECT result.<br /><br />"... but I could replicate it using the more powerfull Chromopainter-Finestructure pipeline."<br /><br />Which I'm being skeptical about. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-39041862849877413772013-07-04T21:42:47.994+02:002013-07-04T21:42:47.994+02:00Not entirely. Saami and Finns did seperate in the ...Not entirely. Saami and Finns did seperate in the Chromopainter-Finestructure as also demonstrated by Vadim Verenich using different method. He had the advantage of having large Saami sample so I couldnt replicate his analysis using ADMIXTURE based tools With so few samples but I could replicate it using the more powerfull Chromopainter-Finestructure pipeline.<br /><br />I called it Middle-East because its frequencies peaks among the Bedouins. They live in the Middle-East not on Sardinia.<br /><br />Maju wrote:<br /><br />"1. The affinity towards Saami you detect is caused by two factors: (a) similarly high score of the NW European affinity component and (b) similarly low score of the Mediterranean affinity ("Middle East"??? - Sardinia is not in the Middle East!!!) affinity component." <br />Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-43992226538803457842013-07-04T18:22:05.325+02:002013-07-04T18:22:05.325+02:00Regarding the images. They are not at Bloggers but...Regarding the images. They are not at Bloggers but directly linked to Google Sites where they appear at full size and can be downloaded and be viewed and zoomed using local viewer and/or editing programs.Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-48580087485696306792013-07-04T17:42:58.002+02:002013-07-04T17:42:58.002+02:00As for the images, these are fine, what really clo...As for the images, these are fine, what really clogs my computer are some (most) of your ChromoPainter heatmap images. These must have brutally high resolutions like 5000x5000 px or something even greater. So when there are several of these my system slows down to a trickle while downloading them, and sometimes it can't cope at all. I'd suggest loading smaller (500x500 or 1000x1000) images and adding a link to the very high resolutions ones below. You can still upload them to Blogger (in a draft post, never published, my avatar is in one of those, for example) or to an image service like Flickr, etc. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-30892242459631571822013-07-04T17:34:30.820+02:002013-07-04T17:34:30.820+02:00Thanks. I'm just wondering how good is K=8 at ...Thanks. I'm just wondering how good is K=8 at cross-validation scores, I feel that more homogeneous regional sub-European components should arise before we can evaluate and that probably requires a K=11 or K=13 level of analysis (but whatever the cross-validation score marks as optimal). But what is clear is that:<br /><br />1. The affinity towards Saami you detect is caused by two factors: (a) similarly high score of the NW European affinity component and (b) similarly low score of the Mediterranean affinity ("Middle East"??? - Sardinia is not in the Middle East!!!) affinity component. <br /><br />2. The affinity towards Basques and Danes (respectively) is caused by two factors: similarly low score of (a) Siberian affinity and (b) Central Asian affinity.<br /><br />3. The orange component, so important and mysterious in La Braña, is shared with mostly peoples of around the Black Sea/Volga area (as well as Orcadians): Chuvash, Turks, Mordvins and Romanians.<br /><br />4. The third component apparent in La Braña is the blue one that also looks from the Black Sea area (Georgians, Lezgins, Turks, etc.)<br /><br />Taken at face value these results La Braña people appear North European with a strong Pontic slant. But as regional sub-European components such as the well known Iberian, Basque, Eastern Mediterranean ones have not been resolved (nor also the difference between the Central-North and Finnish components either), I still find the results a bit inconclusive. One could ask: sure there is greater Northern European affinity than modern SW European peoples but this may well be caused ONLY by more recent Eastern Mediterranean admixture in these, so the result is not really helpful in discerning the pre-Neolithic affinities of La Braña. <br /><br />However it does make them look as pre-Neolithic rather than Neolithic, unless they were exceptionally non-admixed Balcanic peoples, something difficult to argue for but what their composition makes them look to me: people from the Pontic area lacking the Sardinian-Mediterranean component. <br /><br />So there are still many mysteries to unravel in this. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-39599028811175624512013-07-04T09:32:13.621+02:002013-07-04T09:32:13.621+02:00I have made a new post with the results. Sorry for...I have made a new post with the results. Sorry for those huge images. I do not know any other way to present them.<br /><br />http://fennoscandia.blogspot.no/2013/07/la-brana-and-modern-world-variation-ii.htmlAnders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-63288090218414071552013-07-03T14:58:23.617+02:002013-07-03T14:58:23.617+02:00Can't see any new post in your feed, so I can&...Can't see any new post in your feed, so I can't comment. <br /><br />In any case I would strongly suggest you not to upload those hyper-high resolution pictures: in cheap computers like mine at least they really clog the memory up to the point that I had to brute-force-restart my PC twice upon linking to your blog. It'd be much better if you uploaded lower resolution images when possible. I have no problems with most blogs and sites but just very few, particularly yours, being clearly the graphic overload the real problem. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-41987132179013584272013-07-03T14:18:22.407+02:002013-07-03T14:18:22.407+02:00I just did a ADMIXTURE run at K=8 using 4k SNP'...I just did a ADMIXTURE run at K=8 using 4k SNP's from over 2200 worldwide samples in my database. It didn't change the main conclusion from the Chromopainter-Finestructure pipeline analysis. Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-48095769292719950832013-07-02T13:12:16.643+02:002013-07-02T13:12:16.643+02:00I do not know if Dienekes' comparisons produce...I do not know if Dienekes' comparisons produce artifacts or not, they probably do in fact. I'm not particularly fond of his "zombies" either (for example he lets Sardinians cope one of them, even if they are a too homozygous pop., instead he systematically rejects the Basque component, when Basques are clearly much less homozygous than Sardinians or Finns). <br /><br />As I said before, I think that this Braña composite sequence is way too mysterious in too many aspects to draw any conclusions. <br /><br />But whatever the case it is <a href="http://leherensuge.blogspot.com.es/2008/06/larger-samples-better-than-larger.html" rel="nofollow">demonstrated</a> that greater number of SNPs is much less important than better samples:<br /><br /><i>... according to our results a study employing N = 300 subjects and the Affy500K platform offers higher power than a study employs N = 250 subjects and the Ilmn650K platform. This 20% increase in sample size (N = 300 vs. N = 250) provides more power than the 90% increase in the number of SNPs genotyped (286K SNPs vs. 545K SNPs)</i>.Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-59378336325950227682013-07-02T08:28:44.747+02:002013-07-02T08:28:44.747+02:00As I understand it Dienekes DYI calculator based a...As I understand it Dienekes DYI calculator based analysis is based on less than 1000 SNPs and my analysis is based on 183k pruned down to 69k. However in your opinion Dienekes K12b analysis is not "fluked" even it is working with less than 1k SNP's and lack many more northern populations than both my own and Vadin Verenich analysis(?)Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-5967913125092070042013-07-01T21:34:30.476+02:002013-07-01T21:34:30.476+02:00I'm not fighting to persuade you of anything (...I'm not fighting to persuade you of anything (other than maybe try better sampling strategies, if nothing else just for contrast). <br /><br />I really do not care much about what La Braña composite sequence looks like. There's nothing half clear about it: two individual poor quality sequences (maybe totally different from each other?) were merged to create a slightly better quality composite, which looks just weird not matter how you look at it: ADMIXTURE, ChromoPainter or whatever, with African (but NOT NW African) affinities, which now are Asian affinities and then half-Asian, half-African... It is not even clear if these individuals are Epipaleolithic or Neolithic. There's absolutely nothing clear about the La Braña sequence. Sorry. <br /><br />It's you who seem somehow obsessed about persuading me of your ideas and "results". I'm not happy about anything in all that: it simply does not make any sense, so please, let me remain skeptic. It is my right. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-63140531026298960312013-07-01T18:00:51.863+02:002013-07-01T18:00:51.863+02:00Even I only have like 5 Saami samples it is enough...Even I only have like 5 Saami samples it is enought to make inference about minority ancestries in other Fennoscandians.<br /><br />I still can't see you are able to contest any of the inferred ancestries in the project. So far your critism has only been theorical withount any foundation in the actual data.Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-19614674568282380012013-07-01T17:45:52.480+02:002013-07-01T17:45:52.480+02:00More SNPs won't greatly increase the quality o...More SNPs won't greatly increase the quality of your results, instead larger samples, more different samples and different sampling strategies (if only just for contrast) may and probably will enhance the quality of your results. You are basically repeating the same exercise once and again and, sincerely, I don't see any particular merit nor additional information nor contrast exercise in that. If you want to debate all Europe, you must analyze all Europe (and nearby regions), not just that remote "island" of the Far North. And you should use much more balanced samples (otherwise Nordic clusters would phagocyte others' identities).<br /><br />What you do is looking at a few trees and ignoring the forest in which they exist. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-17745121350366953612013-07-01T17:27:11.465+02:002013-07-01T17:27:11.465+02:00I am not sure what this conclusion is based on but...I am not sure what this conclusion is based on but they are welcome to check out the subtle ancestry inference I have infered using the Chromopainter-Finestructure pipeline to see if it is really "fluked" as you define it. It is currently based on 289k SNP's but I have done experiments that make it possible to do even more subtle divisions with a higher number of SNP's.<br /><br />Here is the local Fennoscandian BGA analysis where it is very easy to spot individuals with minority ancestries whatever stated etnicities.<br /><br />http://fennoscandia.blogspot.no/2013/04/updated-local-fennoscandian-analysis.html<br /><br />Maju wrote:<br /><br />"You can't put in the same plane a hard science like Archaeology with soft ones like linguistics or population genetics (yes, genetics is a soft science, scoring similar as economics and well under ecology or agricultural science, not to mention space science). It will depend on the particular facts or claims being compared but in general, what Archaeology produces are facts with high value as evidence while statistical results in population genetics or linguistics are way too often just more or less educated hunches."Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-70887615120954828232013-07-01T16:24:43.200+02:002013-07-01T16:24:43.200+02:00Thats not a problem in Chromopainter-Finestructure...Thats not a problem in Chromopainter-Finestructure pipeline when using superindividuals for all populations except the La Braña. As you can see here the structure make very much sense even the sample sizes are very uneven. The heatmap structure also form nice boxes suggesting the structure is quite right.<br /><br />http://fennoscandia.blogspot.no/2013/06/la-brana-and-saamis-ii.html<br /><br />Maju wrote:<br /><br />"The fact that you heavily load your samples with rare populations from the Far North, often very homozygous, doesn't help but to aggravate the problem."<br />Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-37156145379375821802013-07-01T15:33:24.708+02:002013-07-01T15:33:24.708+02:00You can't put in the same plane a hard science...You can't put in the same plane a hard science like Archaeology with soft ones like linguistics or population genetics (yes, <a href="http://leherensuge.blogspot.com.es/2010/04/genetics-and-biolog-rather-soft.html" rel="nofollow">genetics is a soft science</a>, scoring similar as economics and well under ecology or agricultural science, not to mention space science). It will depend on the particular facts or claims being compared but in general, what Archaeology produces are facts with high value as evidence while statistical results in population genetics or linguistics are way too often just more or less educated hunches. <br /><br />"What of these two seemingly inconsistent results are "fluked"?"<br /><br />A fluke (noun) is "A lucky or improbable occurrence, with the implication that the occurrence could not be repeated". As verb: "To obtain a successful outcome by pure chance". (Wikitionary)<br /><br />I suspect that there's such element in your results because it's very difficult to explain them in any other way, but it's maybe better described as "artifact" than "fluke". Very possibly, the same that the Finnic ultra-homogeneous gene pool causes artifacts in ADMIXTURE and other statistical comparison tools (same for other populations with low heterogozygosity like Ashkenazim, Kalash, etc.), it also happens in ChromoPainter. The fact that you heavily load your samples with rare populations from the Far North, often very homozygous, doesn't help but to aggravate the problem.<br /><br />"As they all weared T-shirts they also should have similar genomes..."<br /><br />Nobody ever claimed that. Serious archaeological reconstruction of cultural realities is not based on single elements but on all kind of evidence. For example most prehistorians agree that Megalithism or Bell Beaker do not represent any single population but wider phenomenons across otherwise locally rooted cultures, similarly the Cardium-Impressed Pottery culture is documented to have some true colonies and also, very often, locally rooted communities which only adopt from CIP the new elements such as pottery and agriculture but not for example the toolkit, typically of local Epipaleolithic roots. <br /><br />But that does not mean that well understood cultural units are meaningless either and in the event of Epipaleolithic colonization of (previously frozen) Northern Europe, which was the point I was arguing, cultural affinity offers us important clues on possible origins of the various populations. Just denying these evidences is not helpful at all, especially if you don't offer alternative clues of the same kind, such as evidence of locally rooted gradual or partial cultural change. Majuhttps://www.blogger.com/profile/12369840391933337204noreply@blogger.comtag:blogger.com,1999:blog-3023805782808412230.post-56567553266957232372013-07-01T09:21:33.872+02:002013-07-01T09:21:33.872+02:00Archeology, linguistic and genetic are all differe...Archeology, linguistic and genetic are all different disciplines, therefore "inconsistencies" between them doesn't invalidate any of their results as they describe very different things. <br /><br />Imagine future archaeologists find what they call the "T-shirt" culture over large distances of Europe. They find T-shirts in Basque, Scandinavians, Finns and Saami burials but when they analyse their physical remains they get a totally inconsistent results with the archeological T-shirts. As they all weared T-shirts they also should have similar genomes, but they didn't.<br /><br />What of these two seemingly inconsistent results are "fluked"? Anders Pålsenhttps://www.blogger.com/profile/13444056522800105747noreply@blogger.com