Feb 24, 2022·edited Feb 24, 2022Liked by Razib Khan
From an explanation point of view for general reader for why a single genome tells so much, I think you missed hammering on the most common baggage a general reader brings to the table. The general reader thinks we know what genes do. So hammering on this NOT being true is critical. I think most readers will still miss this point after reading your post (FWIW). The key insight is genomic ancestry techniques (for the most part) treat the functionality of genes as unknown and not even needed. So what's required is to have a set of baseline genomic guideposts to compare new finds against. That set of guidepost/baseline genomes is what makes a single genome so informative. If we only had a bare bones single Denosivian genome but no baseline to compare it to, it would tell us nothing.
Anyway, enjoying the series as always. And just my two cents of reaction to reading this particular post. If you think there might be something to it, perhaps you can ask readers you know who aren't aware of this, and see if they missed this point or not. And it could help determine what to cover in future.
Great article. And the third paragraph of the opening is one of the most inspired, beautiful paragraphs I've read in a long time -- not the least because it so wonderfully captures what has so fascinated me about this subject since I first sent my sample to the Genographic Project many years ago, and then watched in amazement as the golden age of genetic discovery blossomed. What a ride!
I have a friend with ALS who's had her exome sequenced. Do you think there's any value in doing this sort of commercial testing too? Would there be any realistic way for a lay person like me to compare it against known genetic causes of ALS?
This reminds me of the empirically observed "double descent" phenomenon in deep learning which seems to violate classical statistical intuition about model complexity and size of training sets
From an explanation point of view for general reader for why a single genome tells so much, I think you missed hammering on the most common baggage a general reader brings to the table. The general reader thinks we know what genes do. So hammering on this NOT being true is critical. I think most readers will still miss this point after reading your post (FWIW). The key insight is genomic ancestry techniques (for the most part) treat the functionality of genes as unknown and not even needed. So what's required is to have a set of baseline genomic guideposts to compare new finds against. That set of guidepost/baseline genomes is what makes a single genome so informative. If we only had a bare bones single Denosivian genome but no baseline to compare it to, it would tell us nothing.
Anyway, enjoying the series as always. And just my two cents of reaction to reading this particular post. If you think there might be something to it, perhaps you can ask readers you know who aren't aware of this, and see if they missed this point or not. And it could help determine what to cover in future.
Great article. And the third paragraph of the opening is one of the most inspired, beautiful paragraphs I've read in a long time -- not the least because it so wonderfully captures what has so fascinated me about this subject since I first sent my sample to the Genographic Project many years ago, and then watched in amazement as the golden age of genetic discovery blossomed. What a ride!
I have a friend with ALS who's had her exome sequenced. Do you think there's any value in doing this sort of commercial testing too? Would there be any realistic way for a lay person like me to compare it against known genetic causes of ALS?
This reminds me of the empirically observed "double descent" phenomenon in deep learning which seems to violate classical statistical intuition about model complexity and size of training sets