4 ms·

Holy crap. I was not aware it was possible to map DNA to actual phenotype morphologies. Are there any other examples of out there of this type of work were org

by fsloth 7y ago

Holy crap. I was not aware it was possible to map DNA to actual phenotype morphologies. Are there any other examples of out there of this type of work were organism shape and function can be guestimated to a good accuracy from just DNA?

glofish 7y ago
This is one of those studies that is primarily "belief" based. I for one am not a fan of these - since we can't really disprove it. Let's predict what a Denisovan skeleton looked like based on the ancient DNA of a single specimen... Anyway I don't mind papers like this - they are to show the public that scientists are sometimes working on problems that they can relate too.
dekhn 7y ago
I can't speak to the work's accuracy (I think their PCAs are basically just the 2-3 most common facial variations seen in the wild), it was contoversial when published, but just given the facial similarity of identical twins raised in different environments, one imagines this should be very possible. https://www.pnas.org/content/early/2017/08/29/1711125114 https://www.pnas.org/content/early/2017/08/29/1711125114
- andrewjrhill 7y ago
  I can't believe I have never heard of this before. This is incredible work.
- glofish 7y ago
  One unexplored dimension of this study is that the quality of predictions might get worse and worse as you add more data (individuals).
rossdavidh 7y ago
My understanding is that it is "good accuracy" for purposes such as curiousity about what a Denisovan might have looked like. There was discussion of trying to use it for police purposes (if you have DNA of the perpetrator can you create a sketch of their appearance), and the general consensus was that it was not accurate enough for high-stakes purposes like that. But presumably, someday it may be.
- BurningFrog 7y ago
  Pretty sure you can at least get skin and hair color from DNA now. I'm sure people are throwing CPU peta-cycles on machine learning this as we speak!
  - madhadron 7y ago
    You can get a range of skin colors, since it depends heavily on how much sun an individual is exposed to. Hair color should be pretty straightforward, but, again, depends on the amount of sun and the prevalence of hair dye in the population.

by fsloth 7y ago

glofish 7y ago

This is one of those studies that is primarily "belief" based. I for one am not a fan of these - since we can't really disprove it. Let's predict what a Denisovan skeleton looked like based on the ancient DNA of a single specimen... Anyway I don't mind papers like this - they are to show the public that scientists are sometimes working on problems that they can relate too.

dekhn 7y ago

I can't speak to the work's accuracy (I think their PCAs are basically just the 2-3 most common facial variations seen in the wild), it was contoversial when published, but just given the facial similarity of identical twins raised in different environments, one imagines this should be very possible. https://www.pnas.org/content/early/2017/08/29/1711125114 https://www.pnas.org/content/early/2017/08/29/1711125114

andrewjrhill 7y ago
I can't believe I have never heard of this before. This is incredible work.
glofish 7y ago
One unexplored dimension of this study is that the quality of predictions might get worse and worse as you add more data (individuals).

rossdavidh 7y ago

My understanding is that it is "good accuracy" for purposes such as curiousity about what a Denisovan might have looked like. There was discussion of trying to use it for police purposes (if you have DNA of the perpetrator can you create a sketch of their appearance), and the general consensus was that it was not accurate enough for high-stakes purposes like that. But presumably, someday it may be.

BurningFrog 7y ago
Pretty sure you can at least get skin and hair color from DNA now. I'm sure people are throwing CPU peta-cycles on machine learning this as we speak!
- madhadron 7y ago
  You can get a range of skin colors, since it depends heavily on how much sun an individual is exposed to. Hair color should be pretty straightforward, but, again, depends on the amount of sun and the prevalence of hair dye in the population.