I’m preemptively sorry if this question have already been raised, I don’t feel like reading all the comments.
What am I doing now is reciting the points from Habr (a mostly-coding-oriented site where this article have been translated in Russian and criticized in comments)
The main criticism is against the key point of the article, the graphs of “edit N genes, get M bonus to IQ”.
The arguments are:
There are only a few hundred IQ-related genes, and they’re found through a correlation in over 240k people, so it’s not necessary that you can just edit all these genes to set them on a “more IQ” version in a specific genome, and get maximum IQ. www.southampton.ac.uk/news/2018/03/genes-intelligence.page
By editing more nucleotide sequences, you will only increase the risks of breaking the genome’s reading, replication, expression and other mechanisms, killing the cell. DNA isn’t merely a line of text that encodes proteins, it’s full of commands for enzymes to work with it. As a tip of the iceberg: there are 20k genes that encode 100-1000k proteins in human body, because 1 gene contains several exons (and introns), which, during transcription, are combined in different ways to encode several proteins.
Yeah, a strange hobby I have found for myself: connecting two sides in an internet argument, like a novice chess player copying moves from two grossmeisters to not lose against at least one…
It’s fine, I wouldn’t expect you to read all the comment.
There are only a few hundred IQ-related genes, and they’re found through a correlation in over 240k people, so it’s not necessary that you can just edit all these genes to set them on a “more IQ” version in a specific genome, and get maximum IQ.
www.southampton.ac.uk/news/2018/03/genes-intelligence.page
There have been quite a few more genetic variants linked to IQ since this study was published in 2018. Herasight has the best IQ predictor I know of, which can explain over twice as much variance as the best predictors we had in 2018.
Of course it’s correct that in most cases we aren’t highly certain which one of a cluster of variants is actually causing the effect we observe.
But we don’t need to know. We just need reasonably good odds. Then we can edit variants with the highest combined probability of causality * effect size. That’s how we came up with the graph in the post: we didn’t falsely assume we know which of the variants are causal.
By editing more nucleotide sequences, you will only increase the risks of breaking the genome’s reading, replication, expression and other mechanisms, killing the cell. DNA isn’t merely a line of text that encodes proteins, it’s full of commands for enzymes to work with it. As a tip of the iceberg: there are 20k genes that encode 100-1000k proteins in human body, because 1 gene contains several exons (and introns), which, during transcription, are combined in different ways to encode several proteins.
Yes, that’s true if you’re doing single shot editing, which is one of the limitations for the kind of germline engineering we’re currently pursuing. But the more advanced editing protocols don’t rely on single shot editing. They use iterated CRISPR, as explained in the post, and that in turn uses multiple rounds of editing. In each round, you’re selecting cells that not only received many edits, but didn’t contain any deal-breaker mutations.
So at least in theory, there is no reason why adding more edits would break essential genomic functions.
In practice, there are a lot of headaches here: base editors result in bystander edits, prime editors result in indels, and each cell division carries a certain risk of copying errors and chromosomal abnormalities. None of these are insurmountable, but overcoming them will require a decent bit of engineering.
Do you have a link to the Habr article? I’m curious to read what people are saying.
I’m preemptively sorry if this question have already been raised, I don’t feel like reading all the comments.
What am I doing now is reciting the points from Habr (a mostly-coding-oriented site where this article have been translated in Russian and criticized in comments)
The main criticism is against the key point of the article, the graphs of “edit N genes, get M bonus to IQ”.
The arguments are:
There are only a few hundred IQ-related genes, and they’re found through a correlation in over 240k people, so it’s not necessary that you can just edit all these genes to set them on a “more IQ” version in a specific genome, and get maximum IQ.
www.southampton.ac.uk/news/2018/03/genes-intelligence.page
By editing more nucleotide sequences, you will only increase the risks of breaking the genome’s reading, replication, expression and other mechanisms, killing the cell. DNA isn’t merely a line of text that encodes proteins, it’s full of commands for enzymes to work with it. As a tip of the iceberg: there are 20k genes that encode 100-1000k proteins in human body, because 1 gene contains several exons (and introns), which, during transcription, are combined in different ways to encode several proteins.
Yeah, a strange hobby I have found for myself: connecting two sides in an internet argument, like a novice chess player copying moves from two grossmeisters to not lose against at least one…
It’s fine, I wouldn’t expect you to read all the comment.
There have been quite a few more genetic variants linked to IQ since this study was published in 2018. Herasight has the best IQ predictor I know of, which can explain over twice as much variance as the best predictors we had in 2018.
Of course it’s correct that in most cases we aren’t highly certain which one of a cluster of variants is actually causing the effect we observe.
But we don’t need to know. We just need reasonably good odds. Then we can edit variants with the highest combined probability of causality * effect size. That’s how we came up with the graph in the post: we didn’t falsely assume we know which of the variants are causal.
Yes, that’s true if you’re doing single shot editing, which is one of the limitations for the kind of germline engineering we’re currently pursuing. But the more advanced editing protocols don’t rely on single shot editing. They use iterated CRISPR, as explained in the post, and that in turn uses multiple rounds of editing. In each round, you’re selecting cells that not only received many edits, but didn’t contain any deal-breaker mutations.
So at least in theory, there is no reason why adding more edits would break essential genomic functions.
In practice, there are a lot of headaches here: base editors result in bystander edits, prime editors result in indels, and each cell division carries a certain risk of copying errors and chromosomal abnormalities. None of these are insurmountable, but overcoming them will require a decent bit of engineering.
Do you have a link to the Habr article? I’m curious to read what people are saying.
Thanks for replying!
The link is: https://habr.com/ru/articles/933324/