Comment by dekhn | Hacker News Reader

Comment by dekhn | Hacker News Reader

I, too, have applied for, and then had revoked, access to dbGaP. In my case, I worked for a corporation that wanted to simplify the hosting of this data to appropriately credentialed researchers with the goal of making genomics look more like modern bigtech data processing.

We made friends with the various program managers, did all the necessary due diligence, applied, got approved, downloading hundreds of terabytes of data, and then received a stern note from the leadership at NIH that we had to delete everything. Apparently, some Very Important People at NCBI didn't like Google and really didn't like the idea that we'd have a copy of all this data, even though the intent (I can say authoritatively since I was the founder of the project) was to accelerate scientific research.

It's now hosted on AWS as a public/private dataset.

I try to avoid human genomics or really any controlled-access dataset now. The total value in the data is much smaller than generally claimed, and entrenched interests go out of their way to gatekeep and sort of maintain an "artisanal, bespoke" approach to data processing.

Now, if you want to use this data to make strong conclusions about the relationship between genotypes and human intelligence: best of luck to you, as you'll be grabbing the electrified third rail of American politics and science. Nearly every foray into this area generates controversy, and even intelligent, good-faith researchers end up in a trap of "appearing to be racist, or sexist". With career ending consequences.

I study tardigrades instead now.

> Apparently, some Very Important People at NCBI didn't like Google and really didn't like the idea that we'd have a copy of all this data

You've just improved my faith in government institutions.

Thank goodness Amazon has it instead!

We had a legitimate research use for the data (hosting it for scientists) and it wasn't part of any consumer products. Anyway, I don't work there any more and part of that is I realized that Google can't do research on these sorts of things without getting involved in too much politics.

This actually makes me feel more confident in the system. I as an average citizen, would prefer if an advertising company did not have access to a giant database of genetic information. (I also don’t want Bezos to have access either, for what it’s worth).

I agree with this. Google doesn't need access to terabytes of genetic datasets and I'm glad that the NIH asked them to delete what they had.

Imagine finding a correlation between specific genes and specific spending habits. Have Alphabet spin up a company that does what 23andMe does, offer a cooler, trendier version of the competition to entice adoption, then in the fine print say they’ll mine the data they find in your genes. Those resulting ads are gonna get ugly, quick.

Edit: Although, in the context of medicine, it would be pretty cool if you can find trends in medicinal health that tend to be common with individuals like yourself, and give you warnings should conditions are more likely to arise. I imagine that would involve adding in a different dataset to compliment understanding characteristics of your DNA

There's a ton of research on consumption patterns of those experiencing symptoms of mental illness, and genes can be good indications of predispositions for those illnesses. There's also research on how social media platforms can induce symptoms of mental illness by tweaking the algorithm that shows content to users.

Not only could platforms use that research to optimize engagement/conversions/etc of those with genetic likelihoods of experiencing such symptoms, but unscrupulous platforms can use those genetic predispositions to algorithmically target and induce symptoms of mental illness in the vulnerable. If a platform knows that a user has a predisposition for overeating or depression, they can then try to lead that user into patterns that will make them sick, with the intention of getting them to engage in the desired behavior of clicking, buying, etc.

Never mind that governments are the ones who used similar data sets to murder their own citizens by the millions within living memory. The important thing is that we keep it out of Google's hands!

Do I basically have your position right?

I, and everyone else, get a say in what the government does. It's one of the only institutions that I get a say in, and one of the only institutions that people can audit.

I am not a Google board member, I have no say in what they do. The government does, though, in a limited fashion.

Please don't mistake my desire for Google not to have data with the implicit approval of such datasets existing in the first place.

Also, please remember the millions of people that were starved, enslaved and murdered by the East India Company.

I don't think the East India Company was intentionally committing genocide for the sake of racial/genetic purity.

Offhand, I can't think of any corporations who have done that. There's no profit in it... at least, not unless a government is footing the bill.

Unnecessarily hostile and bad-faith interpretation. We can be more respectful towards each other than that.

That goes both ways. "Google must not be allowed to have access to the data that our taxes paid for" isn't unnecessarily hostile?

If you want to conduct expensive research projects for the exclusive benefit of your own politically-correct cabal, don't ask me to pay for it.

But it's available on AWS. Can't google query it there?

Sure, if a Google researcher applies for and gets access to dbGaP through NCBI, they will get a secret key to decrypt the private parts of the data. But that wouldn't make any sense since this is many many terabytes of data and typically you'd want to convert it to a format that is optimized for data processing.

But that wasn't the point. The point was to have copies of the data preloaded into Google Cloud, preformatted for high performance data processing by a wide range of researchers. One of the highlights of my career was working directly with Jeff Dean and Sanjay Ghemawat on mapreduces that turned FASTQ and BAM files into sstables.

Thanks for your work, even if it got shut down. Good thing the price of gene sequencing continues to fall - eventualy a responsible and open group of academics will start a more accessible database. Illumina's new machine is reportedly $200 a person.[0] At this price 1 million entries in the dataset would cost two hundred million. GWAS starts being useful at a few hundred thousand entries. I honestly don't think 200 m is so much in the grand scheme, given that biotech companies are valued in the tens of billions. The price will also only continue to fall. Science doesn't acquiesce to ideology - it's the other way around.

[0]https://www.bloomberg.com/news/articles/2022-09-29/illumina-...

GWAS is a very limited technique. Most people in genomics still have very simplified models of how complex organismal phenotypes are formed. My goal was to build an embedding between genotype and phenotype that had predictive and generative abilities. Based on everything I've seen, a few tens of millions of genomes for a few hundred highly specific molecular phenotypes is definitely doable if you had all the world's genomics data at your fingertips.

That sounds really interesting. It's a shame it didn't work out.

You made a lot of people happy now.

Wonder if the NIH can ask to delete the data. Shouldnt it be public property since it s tax-funded and open source and all

These are controlled access datasets for a reason- although anonymized, they really do contain a fair amount of PII and NCBI is exceptionally concerned about the potential for abuse. Generally, in terms of weighing the balance between this being "publicly funded, therefore publically accessible" and "publicly funded, available to credentialled researchers through an application process", the NIH has concluded that yes, they can limit access to the data.

For what it's worth I've mentioned elsewhere I no longer work directly with human genomics data because it's such a political hot potato you can't really do any good research.

Reply on news.ycombinator.com