Hacker News Reader

Comment by adsfqwop | Hacker News Reader

As far as I know there is no scientific certainty that CRISPR does not produce off-target mutations. Therefore to use the word "indistinguishable" here seems quite dishonest.

Because of this fact I get the feeling like we are being sold a marketing trick. If the gene editing procedure can produce off-target mutations, then it is not indistinguishable, and to say otherwise is misleading.

"God’s red pencil? CRISPR and the three myths of precise genome editing": http://gmwatch.org/en/news/latest-news/16900-god-s-red-penci...

View on HN

You're right that the precision of Cas9 is not a 100% (off-target effects), but

1) it is still magnitudes more precise than traditional breeding (where you have no idea what is happening and you only select by screening for selected phenotypes); a popular example: https://en.wikipedia.org/wiki/Lenape_potato

2) you will anyway have to verify our intended modification by genome sequencing

Your cited source (GMWatch) is on the same propaganda-level as an Anti-Vaxxer website.

There is also this

Crack in CRISPR Facade after Unanticipated In Vivo Mutations Arise | GEN

https://www.genengnews.com/gen-news-highlights/crack-in-cris...

It's not clear to me why the existence of off-target mutations with CRISPR would be worrying; random mutations naturally happen to genomes all the time, both during normal reproduction, and during selective breeding (which CRISPR could replace). As sibling comments have mentioned, the newly modified genotypes are going to be tested to make sure that any modifications were net-positive.

However, if you are worried about such mutations, you should be asking the question "does CRISPR produce more mutations than a selective breeding program?" Even if mutations are bad, CRISPR might be less bad than what we currently do (for your definition of bad).

If you could write software that could distinguish off-target CRISPR mutations from sequencing data, you'd have quite a valuable tool. Shouldn't be too difficult if indistinguishablity is a declared myth.

That's pretty easy, because exact matching is easy. Just perform sequencing on the new genome, and map the sequences back to the original genome (with edits). If you can't find an exact match, you know something is wrong. You can also look at the statistics: if some sequences are more prevalent than they should be, you know something is wrong.

And how would you differentiate from classical breeding over X decades?

Sequencing has errors - some in the PCR stage so not corrected by greater read depth for short reads (long reads are much more error prone). Genomes are dynamic - they change. You are looking at the progeny or regenerated plant, not the original plant. Exact matching will not get you far.

Yeah, but still it's a pretty easy problem compared to de-novo sequencing, since you have a good reference to map to.

Okay, no. Stop a second. To use an analogy that HN might understand, doing that would be like using those giant javascript frameworks to create a distributed, dynamic website to serve a couple lines of static text.

The answer is blotting. You do Southerns and Northerns and you'll see exactly whether or not you have off-targets. This is super basic stuff, and very reliable.

In your analogy, Southern and Northern would be like using Fortran on a punched card, while high-throughput sequencing is using a modern programming language.

CRISPR doesn't have to be perfect; you simply characterize the modifications that occurred in the line. This is trivial to do with Southern blots. If you get more than one band, you've got more than one insertion. This is a very reliable method, and it would be trivial for any sort of watchdog or oversight body to double check as the probe is simply a portion of the insert sequence.

Reply on news.ycombinator.com