4 ms·

Mitochondrial and Y-chromosomal DNA are not good ways of determining population bottlenecks, since they are inherited exclusively from the mother and father, re

by defen 18y ago

Mitochondrial and Y-chromosomal DNA are not good ways of determining population bottlenecks, since they are inherited exclusively from the mother and father, respectively. The "Two individuals from one race are more different than two random people" myth is a pernicious one that needs to die. The short answer is that you need to look at correlations in DNA, especially of regions under selection. Imagine a single human's DNA as a point in 4-billion dimensional space (where each dimension can take on the value A,C,G, or T). If you do a principal component analysis on the DNA of large numbers of people, you will see clusters, and these clusters correspond to geographic ancestry. http://www.nature.com/ejhg/journal/v16/n12/abs/ejhg2008210a.html http://www.nature.com/ejhg/journal/v16/n12/abs/ejhg2008210a.... "A major East–West gradient from Russian (Moscow) samples to Spanish samples was identified as the first principal component (PC) of the genetic diversity. The second PC identified a North–South gradient from Norway and Sweden to Romania and Spain." In picture form, from that paper in case you don't have access to Nature: http://www.gnxp.com/blog/2008/11/another-genetic-map-of-europe.php http://www.gnxp.com/blog/2008/11/another-genetic-map-of-euro... Now, please tell me again with a straight face that I don't know what I'm "ranting" about, and that DNA in humans is irrelevant.

biohacker42 18y ago
I get paid to do PCA analysis for the biotech industry. Except that PCA is the least I do, PCA is what you learn in a bioinformatics 101 class, there are MUCH better tools then PCA. If two things are noticeably different you don't need PCA, it's redundant. You use PCA, when you're plumbing a huge data set, where things are mostly indistinguishable and you're desperate for ways to separate things out. That's why PCA analysis is so convenient for human DNA, because we're so similar. Look at this image: http://scienceblogs.com/gnxp/genmapeuropeGermany.jpg http://scienceblogs.com/gnxp/genmapeuropeGermany.jpg Are you telling me that the Germans from the blue part of Dresden bordering Munich and the Germans from the green part of Munich bordering Dresden are super duper different? That picture sure makes it looks like Hans living in the exurbs of Munich and Frans from the exurbs of Dresden are two different species. It looks that way if you don't have a fucking clue about what PCA is and how it works and what DNA is and how it works. And two individuals from the same race are more different then two large groups of people from the opposite end of the earth, is not a myth. It is a fact. It is a statistical fact, the math doesn't care what you think and your grasp of stats and genomics is not nearly as tight as you imagine it to be. No one imagines DNA as 4-billion dimension space, you're full shit. And if you start looking at what ever regions you're fantasizing about, and apply a bunch more conditions, you can prove anything you want. That's not science it's charlatanry. In conclusion, you do not know what you're ranting about, so stop embarrassing yourself in front of biotech engineers on the internet.
- defen 18y ago
  I intentionally simplified because I didn't know you were a biotech engineer. Also I'm impressed that you were able to divine how much I know about statistics from my few short posts here. So you're telling me that if I do PCA on the covariance matrix for many loci, I'm not going to be able to correlate DNA with geographic ancestry? How do you explain this: http://www.sciencemag.org/cgi/content/abstract/298/5602/2381 http://www.sciencemag.org/cgi/content/abstract/298/5602/2381 I will grant you one point - "Within-population differences among individuals account for 93 to 95% of genetic variation; differences among major groups constitute only 3 to 5%." However, if you read the paper you'll find that they were able to identify clustering along geographic lines, and could place individuals with a discrepancy rate of 0.14%.
  - biohacker42 18y ago
    Obviously you can correlate DNA with geographic location. But in-population differences of individuals are bigger then inter-population group differences because all the populations are so similar. And because the differences between the populations are so small you'll only be able to measure them, when you measure very, very, very narrow things. Like PCA does. For example you could measure skin color. The amount of melanin deposited in your skin is controlled by one gene. So that's about as narrow a measure as you can have. But can you draw conclusions about broader characteristics in phenotype? Nope. That's why the whole race concept has fallen out of favor, not because it's a lefty conspiracy to avoid hurting people's feeling. It's not that, it's science. The same goes for IQ and societal success. Can you correlate education levels with a phenotype or genotype? If you could, scientists would have already done it, and they would be famous. Parents would be screening their embryos. You could do it and be famous. Again, it's not some lefty conspiracy, it's just not possible. Oh we do have genes that seem to result in more scientists and engineers, the Ashkenazi have them, but they correlate with neural disease. And the Ashkenazi don't have a higher incidence of CEOs and politicians and musicians. And who's on top of the societal ladder in western society? Politicians and CEOs or us geeks? And who has a higher reproductive success, musicians or engineers? That's why people talking about poverty, race and IQ, are like people talking about Monster audio cables. Your assertions about IQ and Mexicans are not evident from the real world genomic data. But if you're prejudiced you can try and twist the facts to match your prejudices.

by defen 18y ago

biohacker42 18y ago

I get paid to do PCA analysis for the biotech industry. Except that PCA is the least I do, PCA is what you learn in a bioinformatics 101 class, there are MUCH better tools then PCA. If two things are noticeably different you don't need PCA, it's redundant. You use PCA, when you're plumbing a huge data set, where things are mostly indistinguishable and you're desperate for ways to separate things out. That's why PCA analysis is so convenient for human DNA, because we're so similar. Look at this image: http://scienceblogs.com/gnxp/genmapeuropeGermany.jpg http://scienceblogs.com/gnxp/genmapeuropeGermany.jpg Are you telling me that the Germans from the blue part of Dresden bordering Munich and the Germans from the green part of Munich bordering Dresden are super duper different? That picture sure makes it looks like Hans living in the exurbs of Munich and Frans from the exurbs of Dresden are two different species. It looks that way if you don't have a fucking clue about what PCA is and how it works and what DNA is and how it works. And two individuals from the same race are more different then two large groups of people from the opposite end of the earth, is not a myth. It is a fact. It is a statistical fact, the math doesn't care what you think and your grasp of stats and genomics is not nearly as tight as you imagine it to be. No one imagines DNA as 4-billion dimension space, you're full shit. And if you start looking at what ever regions you're fantasizing about, and apply a bunch more conditions, you can prove anything you want. That's not science it's charlatanry. In conclusion, you do not know what you're ranting about, so stop embarrassing yourself in front of biotech engineers on the internet.

defen 18y ago
I intentionally simplified because I didn't know you were a biotech engineer. Also I'm impressed that you were able to divine how much I know about statistics from my few short posts here. So you're telling me that if I do PCA on the covariance matrix for many loci, I'm not going to be able to correlate DNA with geographic ancestry? How do you explain this: http://www.sciencemag.org/cgi/content/abstract/298/5602/2381 http://www.sciencemag.org/cgi/content/abstract/298/5602/2381 I will grant you one point - "Within-population differences among individuals account for 93 to 95% of genetic variation; differences among major groups constitute only 3 to 5%." However, if you read the paper you'll find that they were able to identify clustering along geographic lines, and could place individuals with a discrepancy rate of 0.14%.
- biohacker42 18y ago
  Obviously you can correlate DNA with geographic location. But in-population differences of individuals are bigger then inter-population group differences because all the populations are so similar. And because the differences between the populations are so small you'll only be able to measure them, when you measure very, very, very narrow things. Like PCA does. For example you could measure skin color. The amount of melanin deposited in your skin is controlled by one gene. So that's about as narrow a measure as you can have. But can you draw conclusions about broader characteristics in phenotype? Nope. That's why the whole race concept has fallen out of favor, not because it's a lefty conspiracy to avoid hurting people's feeling. It's not that, it's science. The same goes for IQ and societal success. Can you correlate education levels with a phenotype or genotype? If you could, scientists would have already done it, and they would be famous. Parents would be screening their embryos. You could do it and be famous. Again, it's not some lefty conspiracy, it's just not possible. Oh we do have genes that seem to result in more scientists and engineers, the Ashkenazi have them, but they correlate with neural disease. And the Ashkenazi don't have a higher incidence of CEOs and politicians and musicians. And who's on top of the societal ladder in western society? Politicians and CEOs or us geeks? And who has a higher reproductive success, musicians or engineers? That's why people talking about poverty, race and IQ, are like people talking about Monster audio cables. Your assertions about IQ and Mexicans are not evident from the real world genomic data. But if you're prejudiced you can try and twist the facts to match your prejudices.