Hacker News Reader

Comment by rflrob | Hacker News Reader

While direct editing of bases will be a really useful tool, I think it's worth pointing out to any alarmists about human editing that our ability to make directed changes far outpaces our knowledge of what changes to make (and our ability to make reasonable guesses about how safe those changes are). Although there are a relative handful of diseases that have a single, protein coding change that's responsible (sickle cell, cystic fibrosis, etc), most diseases are due to multiple variants, many of which are non-coding, interacting with environmental factors and chance (diabetes, obesity, depression). These kinds of tools are incredibly useful in the lab for making advances in understanding disease, but we're a long way off from widespread clinical use, if that is even in the cards.

View on HN

CRISPR has arrived at a time when we're also making huge advances in bioinformatics and machine learning. When you've got a reliable tool for editing DNA, many biology problems reduce to software problems. The big risks and big opportunities reflect the synergy of multiple technologies. CRISPR alone is a useful curio, but CRISPR plus high-throughput sequencing plus shared genome databases plus ASIC-based machine learning is something else entirely.

It's also worth considering that conventional western standards of ethics and acceptable risk won't necessarily be a constraining factor on the development of genetic engineering over the next few decades.

I'm with you on single base edits being mostly not applicable to humans clinically, but...

"our ability to make directed changes far outpaces our knowledge of what changes to make"

On this point I couldn't disagree more. It's the opposite: our knowledge of what changes to make is years ahead of being able to make those changes. The polygenic scores available for complex traits now, though crude, are within an order of magnitude of what correlations will ever be possible.

For example, the best PGS for educational achievement explains somewhere around ~10% of variance, which corresponds to a correlation of ~32%, and the total heritability is only something like ~75%, meaning the best possible correlation would be ~87%. Anything you can do with a 87% correlation, you can also do with a 32% correlation, and get a proportionately weaker effect.

Of course, there are still many unanswered questions about which SNPs are really causal, and the tradeoffs associated with editing them, but you don't really _need_ this to make edits that are much more likely to help than hurt. As long as you stay within the bounds of what could have happened naturally by chance, the risk really isn't big.

The real problem is that you have to make a _lot_ of edits, and the editing technologies aren't ready for that. I don't know an exact number, but you might be looking at 100s of edits in a single embryo to do anything noticeable. Traditional double-stranded-break CRISPR can't even do 1 edit without usually causing some major damage. Base editing looks better, but still can only do a small proportion of all possible edits.

In conclusion, if you could print an arbitrary human genome, you'd see some pretty sci-fi stuff even today. But we're pretty far off that.

Clinical use is a when not an if http://www.nature.com/news/crispr-fixes-disease-gene-in-viab...

Clinical use is definitely coming, no argument there. What I am arguing is that most people don't have any medical issues that are worth the risks, expense, and ethical considerations of gene editing. I'd also doubt that will change much in the lifetime of anyone alive today.

> our ability to make directed changes far outpaces our knowledge of what changes to make (and our ability to make reasonable guesses about how safe those changes are)

But won't ability to make changes lead to better knowledge of wehat these changes do? E.g. it would make it easier to perform experiments (on non-human species, one would hope) where changes are made and the results studied.

One mistake, and it could affect all of humanity in short order. Gene drives, in particular, highlight the potential potency of our ignorance:

https://en.m.wikipedia.org/wiki/Gene_drive

The potential for good is enormous when it comes to Gene editing. But it's not risk free. (Just like nuclear energy.)

I'm more terrified of bioweapons targeted at people with some specific gene and attached to a flu virus. Genocide at a push of a button, orders of magnitude cheaper than a nuclear weapon, no materials other than standard lab equipment needed. Bond villains, rejoice!

Any reason to hope we won't be able to do this for another hundred years? Because CRISPR seems awfully close.

How would you prevent natural mutations from modifying or disabling the target mechanism? If you can't guarantee its integrity, the descendants of your air-born killer virus will be back for you in short order. That, rather than inability to do it, is what will hopefully prevent it.

That's only applicable for relatively rational state actors not insane extremists who believe such terrible events are required as part of their faith

So many reasons!

  * In order to get CRISPR to fit inside a virus you have to remove all the viral DNA, so it can't reproduce. 
  * Even if you could make such a virus, there would be enormous selective pressure to ditch the CRISPR proteins, and there's no pressure maintaining the CRISPR or guide RNA sequence, so it would be inactive pretty quickly
  * Race isn't well-defined genetically so it might not even be possible to target a group of interest
  * Bombs are cheap and a proven technology

Reply on news.ycombinator.com