Hacker News Reader

Hacker News

Comment by danharaj | Hacker News Reader

Parent Full thread

danharaj

This is why you should be very skeptical of anyone who claims transparency between genetics and complex phenotypes.

View on HN

Indeed. The whole "One Gene, One Protein [One Phenotype]" [1] dogma went out quite some time ago.

... But people and genetic testing companies seem to want to keep it going.

[1] - https://embryo.asu.edu/pages/george-w-beadles-one-gene-one-e...

We will eventually find that there's the equivalent of JPEG encoding (plus SSL and ECC) and that today we're doing the equivalent of someone with no knowledge of SSL or JPEG staring at a tcpdump trace of a browser loading an image over https, trying to make sense of the resulting hex dump.

Can you elaborate - I’ve always been fascinated hearing CS interpretations of biological processes.

DNA methylation acts as a form of error correction analogous to parity.

If genes are the data, then methylation is more like an additional piece of metadata directing the cell (potentially different across cells) how to use the data. One possible use for this metadata is potentially error correction, but it does much more than that.

And alternate splicing is a form of compression, where exons store the repetitive parts of genes only once, to be combined to make many proteins.

Trying to figure out how a single compressed/encrypted buffer works given only it and its resulting image is certainly an unwinnable task, but increase the sample size to a few million or billion and I don't see why the algorithm couldn't eventually be cracked.

Great analogy. This might be the case with some of physics, too.

> But people and genetic testing companies seem to want to keep it going.

You can still get alot of mileage out of a simplified, inaccurate model. If you just say to yourself, "well, this whole gene thing is hopelessly complex", then you'll never get anywhere outside of a lab or personal project. Applied biology especially, and medicine in particular, seems to require grossly simplifying assumptions in order to make any problem tractable.

I would think that would be understood by people on HN. One of the biggest impediments to smart people starting companies is an inability to set aside the inherent, irreducible complexity that they, alone, see and appreciate. The people who succeed in business tend to be those ignorant of the complexity[1], or those able to come to terms with the reality that any practical, marketable solution is necessarily built on an inaccurate model of the world and is doomed to eventually crumble for its inherent flaws.

The old adage that ignorance is bliss is more insightful and meaningful than people give it credit for. Knowledge, especially epistemic knowledge, can be debilitating.

[1] Which doesn't imply being stupid or incurious. Willful ignorance seems like a potentially useful strategy.

You are basically saying "lying to people can make you some money". Indeed…

That doesn't really follow. We don't need to have a complete map genotype to phenotype in order to have an understanding of the weight of genetic component of certain traits. Just because we don't know exactly which genes cause a certain hair color doesn't mean we can't guess which parents are more likely to have children with that hair color. The same goes for behavior that can be measured and predicted. As long as that prediction is more accurate than a random guess, the prediction has value.

If you drop an apple & observe that it falls, you get some predictive value about the nature of gravity. But that's a far cry from computing the orbital dynamics of a spacecraft. Parent's comment on "transparency" in the mapping from genotype to phenotype was more akin to the latter -- and their point stands.

You can read their comment history to contextualize their angle.

We sent astronauts to the moon with Newtonian physics. No need for relativity. This is more akin to relativity in that it expands our domain knowledge and explanatory power but the old model already does most of the work.

Do you not think studying gravity lead to predictive value before we reached a point of being able to compute orbital dynamics of a spacecraft?

The genetic basis of complex traits was the biggest open problem in science (IMO) long before this discovery came along.

Which is why I have zero interest in things like 23 and me. I worked in a genetics lab and the longer I worked there, the more I realized how little we understand.

Sure, we don't understand much detail but we know a whole lot of things like "you're x% more likely to get y disease", and if y is something that can be monitored or tested for, this can literally save your life.

BRCA and heart disease come to mind.

Reply on news.ycombinator.com