That study has done in reality what the AlphaGenome Atlas does in fiction, but instead for a human they have done it for one of the simplest viruses.
So they have fuzzed the virus by mutating one by one each position of its DNA.
And various dedicated AI models all made poor predictions of the results of that experiment, which casts doubts about the value of the AlphaGenome predictive map.
A virus is much simpler than a human, but even for that simple virus the effects of most of the mutations could not be predicted. A half of the mutations had harmful effects, and for a half of those it is unknown for now why they were harmful.
For a human the uncertainty about the effects of a mutation will be far greater than for one of the simplest viruses.
> And various dedicated AI models all made poor predictions of the results of that experiment, which casts doubts about the value of the AlphaGenome predictive map
I would not group AlphaGenome into the pile of failed predictions of other models. AlphaGenome deserves to get evaluated based off its own merits.
Yep. Sequence-to-function models are still very limited. AlphaGenome Atlas, despite the flashy branding, is unlikely to provide significant benefit to researchers.
Imagine cellular activity as an industry zone, its not just what you can or can not make, its also 'for what concentrations of chemical species, what transcription rates should be used' so apart from the discrete Mendelian aspects (like what eye color or what have you) there is also a concensus sequence and deviations from consensus. They mention the dataset captures non-coding DNA, which should imply promoter sequences. Will it be possible to query the atlas for joint probabilities of promoter and putative target protein occurence in human genomes?
Personalized medicine could never credibly take off as long as promoter sequences were excised before sequencing!
23andMe and similar companies don't transcribe your entire genome because that would cost way more than they charge you. They just sample a few tiny sections of it.
Thank you for the recommendation! I've been on a lookout for an affordable full sequence and 400$ is very reasonable. Very excited to just have my DNA on my own computer - this feels like the good kind of sci-fi :)
Probably not any 23andMe haven't already told you about. They test a limited set of SNPs, balancing between ones thought useful for genealogy, ones useful for ethnicity estimates and ones thought useful for health-related things (the latter they would like to make their main selling point, the two former are really all commercial DNA services' bread and butter).
It's unlikely that they would luck into testing some unknown SNP which turned out to be relevant for disease.
23andMe tests SNP's (single nucleotides) that are inferred to be significant in protein function/epigenitics.
Those SNP's i believe are testd from primers
so what 23andMe does is specifically on the back of previous research and afaik their data isnt technically clinically significant as most findings need confirmation or more tests.
Why not? There is no logical reason as to why this would not work, IF it works in the first place, which I don't know. In theory the problem space here is finite, so there is of course a way to predict everything. Whether this is the case right now - who knows; I probably don't think it is currently ready. But eventually it will be. And it should not be in the hands of private companies.
So it sounds like you're coming to this with very little knowledge about biology. I encourage reading up on modern challenges in pathogenic prediction, especially with regards to SNPs: on their own, with the exception of a few diseases, individual SNP predictions are meaningless in terms of actual pathogenicity.
Alpha Fold has continued to impact the field of protein networks, but I do hear that not every one of the deep learning biology models from Google/Deep Mind and others have made equivalent impact or had as lasting relevance in their respective domains..some have performed more poorly than other available models. I'd love to learn more about this, but this has mostly come from little snips of conversations here and there, in person and online, but I haven't seen anything comprehensive in terms of evaluating their impacts overall
People are upvoting this because it has the “Alpha______” prefix. Meanwhile, everyone in the field of genomics knows that AlphaGenome provides essentially zero improvements over the previous SOTA, Borzoi…
> a database that predicts the effects of every possible single nucleotide variant in the human genome. We used the AlphaGenome AI model to pre-calculate the regulatory impact of all 9 billion single-letter genetic changes, resulting in a massive, 1-petabyte dataset.
This is for a database, no? While Borzoi is a model?
> Here, we introduce Borzoi, a model that learns to predict cell-type-specific and tissue-specific RNA-seq coverage from DNA sequence.
> Meanwhile, everyone in the field of genomics knows that AlphaGenome provides essentially zero improvements over the previous SOTA, Borzoi…
Can you elaborate on this? I'm confused why Google would build something that provides zero improvements over SOTA, Borzoi... as you mention. I'm not familiar with this field, just curious.
Speaking as somebody who has worked within Google Research before: the researchers are under tremendous pressure to publish SOTA and sometimes they juice their results a bit to look competitive when they can't match. This is not uncommon in the field- it's remarkably easy to edit a paper to make yourself look good by omitting information.
One of the most egregious cases of this, in my opinion, is only publishing metrics that cover part of the confusion matrix. “The false-negative rate? That could not possibly matter for a variant effect prediction model; why would we include that in the paper?” Example: AlphaMissense.
The use of an exact quote in an ungrammatical fashion is a bit of a language model smell. I can’t help but be reminded of the purely nonsensical AI interview answers. “It’s a pleasure to meet you, Chick Bongo”
Creating an account 12 minutes ago (from the time of this posting) to comment on how another comment seems to you like a "language model smell" is itself, a language model smell, or a scammer.
Please stop accusing or hinting at others being a language model or bot. Not only is it a dumb waste of time, it's wrong in this instance and you are not only going to continue to be wrong but you have no way to prove or demonstrate that any single post comes from a bot nor the ability to do anything about it if you did in fact believe some comment to be attributed to a bot.
Is this just Google precomputing Alpha genome values - which were already accessible via API and making them available as another API (presumably more broadly)? Or is there actually new information?
This has Demis written all over it. There is a great video of him with AlphaFold chatting with the team about releasing some results, and he asked something like “what if we just do them all?”
I don’t think Demis played a big role in this. It was mainly Ziga Avsec who developed Enformer (the first actually decent sequence-to-function model), and then AlphaGenome.
This has nothing to do with AlphaFold at all- not sure if you were implying that.
(the scientific contriution is welcome, but it's not particularly significant)
I saw a really interesting talk by Katie Pollard at ISMB this year about the limitations of variant prediction.
The gist was, can existing variation provide enough context to infer impact of variation? The answer seemed to be no.
Kind of like how frontier LLMs need to ingest larger and large amounts of text to advance. We are going to need to leverage comparative data from other species, and likely tremendous amounts of laboratory mutagenesis experiments to actually make headway on variant prediction. Nature, as it stands, just doesn't have enough human variation.
That's an interesting statement: "The gist was, can existing variation provide enough context to infer impact of variation? The answer seemed to be no."
Is this saying that if we were to sequence every human being on the planet, we'd still be unable to explain some phenotype differences caused by variation simply becase there aren't enough humans/enough variation? Interesting, as that's the first time I've heard that claim, and it would suggest that we spend our time working on mechanistic models of variant to phenotype.
Yes, and genetic variant generally do not act in isolation. We currently focus on the small additive effects of variants because we can with small sample sizes—and 1 million humans is marginal using a GWAS cohort to dive into epistasis. But these interaction effects among variants are critical. Now almost completely deprecated.
It's "shareholder-linked" (probably not "driven"), but yes, upvoted. Nonetheless, even when with grave faults in the path, Google has managed to give us marvellous, paramount free resources (Google Street, Google Museum...), and would have given more (Google Books - if the negotiation had succeeded). I am grateful.
Google Streetview/maps isn't even the best mapping/'streetview' tool at this point. Filled with ads and bugs, some major bugs have been there for 10+ years.
For maps there's quite a few, just anything that doesn't have the amount of ads beats it. For Streetview apple lookaround is a better experience if it's mapped well in your area.
I believe Google has been trying to improve their Streetview in coverage as apple has become a concurrent, but it still has the same bugs people have been complaining about for years (like going backward when pressing forward type of bad). If I remember correctly the bugs have been 'closed' or something before just to include it in the same slightly different way.
And which public enterprise has undertaken those projects?
Looking at current hot problems, and exemplary of the situation: which public entities have endeavoured to backup the most important channels and videos from YouTube, the current historic video archive?
Why would I be grateful?! Because somebody did create those resources and made them available! And they did to common good (if anybody refused the expression "for" common good).
What's your alternative? Executive fiat? Mob rule?
Deepmind is objectively good. These endeavors would always cost money. Better private capital than state, no one has to give Google anything. The state is not so generous.
Every single serious theorist of democracy has recognized this.
Democracy is about rule of law and institutions, not simply the mob.
Tyranny of the majority is the paramount concern of all democracies. Nevertheless, institutions can fail.
The idea that democracy always leads to desirable outcomes is evidently false by historical record.
I just don't understand people who think the state is better than Google. Google pretty much works. Google does not have police powers. Google cant surround your car and point guns at your family. Most of the bad things about Google are actually forced upon them by the state. Who could have even a passing knowledge of history and think states are somehow safer than corporations?
Take the United States. Should Donald Trump get a say in how to run Google? Maybe you like him, so how about the guy that comes after him, or the next one after that?
The American mob seems intent on leading their country into the gutter. Good thing (despite what the mob has thrown at it) that the United States still has strong institutions and strong markets. Without Silicon Valley, America would have already lost its place on the world stage.
But please lets hear more indictments from a population who can't even agree with itself on the issues already under the government's purview. Now let's set a precedent for even more state interference in the market, that way when your party (whatever side you are on) chokes the next future election away the next nut you don't agree with has even more power? Is that your plan?
> Should Donald Trump get a say in how to run Google?
Trump does have a say in how to run google, in a very real sense.
> The American mob seems intent on leading their country into the gutter. Good thing (despite what the mob has thrown at it) that the United States still has strong institutions and strong markets. Without Silicon Valley, America would have already lost its place on the world stage.
American elites are more than happy to follow basically whatever Trump wants them to do with the knowledge that they'll get preferential treatment by the law and win all the no-bid contracts he can award. But ask yourself if this would be any different if Trump were head of a large company under a smaller regulatory state. Seems pretty obvious that not much would change.
If by "the mob" you mean, a small cohort of wealthy, powerful, unscrupulous people, those people who make decisions that the rest of us have no say in, then I agree, that is a problem.
> But please lets hear more indictments from a population who can't even agree with itself on the issues already under the government's purview. Now let's set a precedent for even more state interference in the market, that way when your party (whatever side you are on) chokes the next future election away the next nut you don't agree with has even more power? Is that your plan?
In a smaller state, Trump would have continued to be an executive and the shape of his impact on the world might be slightly different, but would be no less oppressive. Ask tenants in NYC in the 1980s if they think a Donald Trump with smaller regulatory oversight who stayed in business would be better or worse for our country.
Do you think our country flips parties every year because our government does a good job of reflecting the will of the people?
> Trump does have a say in how to run google, in a very real sense.
Indeed, as does every President or government. But surely, you have noticed and take seriously that he has acted with greater executive fiat? That by American standards, while he may not be totally alien, he is not "normal"? The question then is whether we want to accelerate along this trend. What is interesting is many populists, whether on the right or left, seem to answer in the affirmative, confident giving their preferred president increased power is the right trade to make. I simply ask, what happens when the government flips, as it always does? Then maybe you will come to regret your position.
> American elites are more than happy to follow basically whatever Trump wants them to do with the knowledge that they'll get preferential treatment by the law and win all the no-bid contracts he can award. But ask yourself if this would be any different if Trump were head of a large company under a smaller regulatory state. Seems pretty obvious that not much would change.
The first part of this is ahistorical. American elites always line up behind the President. If Kamala had won Zuckerberg would be out front with DEI or w/e. The big companies, banks, Silicon Valley, always shower money and gifts on the incumbent. The government operates in many ways as a big public/private partnership. Again, the danger with Trump is the degree, not necessarily a wholesale difference. Some people use this to say it doesn't matter, makes no difference. I'm not one of those people. I believe at a certain point, tip the degree too far, and a difference appears.
As for the second part, democracy allows the people to correct for their mistake. The elites did not want Trump, and never have. He came to power after descending an escalator, elites on both sides mocked him. Elites wanted to repeat Bush v Clinton. Trump may be wealthy, but he is not an "elite", he is new money, and has basically always been a joke among the upper echelons of power in New York City. The elites are nice to him because they are forced to deal with who the mob elected. To not do so is shareholder suicide. Everyone knows he is insane. They simply cannot say it out loud. And why would they? What's with people who simultaneously think Google has too much power and yet want them to be out in front fighting Trump? It's not their job. Why did the American people elect him? What is Google supposed to do? Ignore the President of the United States? Tell the American people who to vote for?
> If by "the mob" you mean, a small cohort of wealthy, powerful, unscrupulous people, those people who make decisions that the rest of us have no say in, then I agree, that is a problem.
No I mean the ~100,000 average Americans who actually make a difference in electing presidents, in places like Wisconsin, where the state is sufficiently purple and there is a small "swing vote" that is often well under a million people. When are they going to be held responsible for their mistake? Why is Google my enemy? I can stop using Google. Google provides me excellent services at a reasonable price (often $0 - I understand I give them data but who cares, really only the government would care about my individual data, and I know Google simply aggregates it to "subsidize" the numerous free services I get). So, a pretty fair deal, and one I can not enter to if I actually wish. On the other hand, what does someone in Wisconsin give me? I can't stop someone in Wisconsin from voting against my interests. Is Google supposed to be the tyranny here? Extend this same logic to any state on the wrong side of your preferred electoral divide. Like many Americans, I really have no say in Presidential elections, because I live in a populous, blue state. My vote doesn't matter, my electoral votes are already priced in. Meanwhile people in red flyover states get the same number of senators as my state, which has a multiple of their population. This makes corporations my enemy? Did Google sponsor the Connecticut Compromise?
> In a smaller state, Trump would have continued to be an executive and the shape of his impact on the world might be slightly different, but would be no less oppressive. Ask tenants in NYC in the 1980s if they think a Donald Trump with smaller regulatory oversight who stayed in business would be better or worse for our country.
Let's not move the goalposts. I never said I was against all government regulation. If you want to regulate Google, then fine. Make a sensible proposal. What is not serious nor sensible is saying "Google is evil" or "Google should not exist".
> Do you think our country flips parties every year because our government does a good job of reflecting the will of the people?
You need to not think only of the current case relative to your supposed best case, but also a worse case. What is the will of the people? Who decides? How? It seems obvious to me the government being defective is somewhat by design (separation of powers) because the founders knew as many people know now, that America is very large, very divided, and very dangerous. A federal government that doesn't give some of the voters 100% of what they want every four years is a lot better than another civil war. Yes things could be better. They could also be a lot worse.
I tend to think people who say "Google is evil" have a rather limited view of the downside risks that occur when humans congregate. Google does not even begin to scratch the surface of the historical depravities we have visited upon each other.
Have you been to states that predominantly vote against your particular preferred political party? You will find "the corporations" are not your enemy. It's what it always has been, all the time, throughout every historical age. The problem of civilization is other people. You will not get around that by trying to treat corporations as some special case.
Propose sensible regulations. Maybe even a way to reform the electoral college that has a chance of passing. I am interested in these ideas. Innovation can happen in technology and in government. That's why I like this forum. But too often interesting conversations are drowned out by "Google is evil" as if this is useful. I would love to hear more interesting solutions, and less hyperbole. Thanks for the pushback.
> Market reward is the best proxy for good we have.
Only if the goal is profit. All else be damned.
> What's your alternative? Executive fiat? Mob rule?
It is always bewildering to me that people see the world being raped by corporations and their negative externality, and they think that this is the pinnacle that humanity can achieve.
Nothing else can be better, if all your options are "mob rule" or "executive fiat".
I fancy myself a misanthrope, but damn, next to people in this forum I am probably the last humanist walking through the ruins of civilization.
> Deepmind is objectively good.
Google is objectively bad, and the world would be a much nicer place if they ceased to exist.
> It is always bewildering to me that people see the world being raped by corporations and their negative externality, and they think that this is the pinnacle that humanity can achieve.
> I fancy myself a misanthrope, but damn, next to people in this forum I am probably the last humanist walking through the ruins of civilization.
A humanist who uses a term of sexual violence so cavalierly?
If you are in the United States and you want a good civic lesson in state power, you should go to your nearest metropolitan criminal court. Maybe you will have a more accurate view of what government means then. At the very least, you will likely learn to banish such a casual invocation of that word from your lexicon.
> Nothing else can be better, if all your options are "mob rule" or "executive fiat".
What we have now is the middle ground between king and mob. It took centuries and much carnage to achieve. Does the famous Churchill quote really need to keep being repeated? Do you have an alternative proposal? Or maybe you have a sensible proposal that falls within the established democratic bounds? Or perhaps instead by "would be a much nicer place if they ceased to exist" we should simply infer a generalized whining rather than any specific policy prescription?
What's your actual idea, do you have one?
Maybe if you endeavored to communicate without self-satisfied hyperbole you would find we actually have reason to agree with each other.
Executive fiat, given the right executive body, and undergirded by healthy autarkies or other cell-type self-direction, worked well in most locales for most of the years humans have been humans. Certain flashy civilizations, Sumer, Greece, have tended to draw the spotlight incommensurately with their population sizes. But even these functioned much more like what I describe above than anything modern.
I found this one of the more entertaining formulations of "back to the land" style sentiments that I have ever come across here, so the downvotes you are getting aside, I thank you for the humor.
Alphafold or Earthquake alerts for example spring to mind.
You could argue about comprehensive "free" services (search, Gmail, YouTube, docs, etc) and "free" browsers and "free" operating systems (yeah plural) being net positives in the world but I know that many people view these as negatives as they (rightly or wrongly) view them as vehicles to sell ads or collect data, but I bet you 100 bucks the typical person on the street using their android phone to search Google, watch YouTube, check their Gmail, write a doc, view photos etc doesn't give a shit about that and values these things - google has ended up as a monopoly as these things are actually pretty useful to people and thus valued, even if we in the tech world disagree with the adtech and data aspects. This monopoly was not a fait acompli or pure market abuse - Nokia, MS, flikr, Hotmail, yahoo etc were all incumbent and market dominant but are now irrelevant because the Google products were viewed as better by end users.
I don't want to drag the discourse away from this achievement, but I hate how this is announced with blatant corporate advertising (our internal model, here are the benchmarks, gpt astra TM yours now for the low low price of £200pcm). I just didn't think Navier-Stokes falling would be sponsored by McDonald's.
Still. I am crying right now. Navier-Stokes is solved.
There are AI labs headed by marketing CEOs that fake metrics, have an utter disregard for humanity and make up "AGI is imminent" propaganda for their IPO, and then there are AI labs headed by actual scientists that do actual science for humanity without constantly trying to put themselves into the spotlight.
the Google DeepMind PR team can't catch a break, shame it makes no sense to me - if someone's in the field maybe they could explain to the rest of us if this is a big deal, just a PR move, or nah?
I think orliesaurus meant the PR team is unable to rest because the rest of Deepmind is so prolific and keep coming out with new things for them to announce
not my field so I can't judge how useful it is but assuming this data can be used for drug discovery and their ToS limiting to non-commercial use only. Does DeepMind plan on selling this data to pharmaceutical companies?
You deleted the verbatim comment after it was pointed out. One was in thread and one was a stand alone comment. I was curious which was the throwaway so clicked into both profiles
Videos. [2] is for the scientists to start using AlphaGenome Atlas from AntiGravity.
1. https://www.youtube.com/watch?v=U0aToL5C-bQ
2. https://www.youtube.com/watch?v=b2qw3rDNX0Q
In another HN thread about this AlphaGenome Atlas, someone has posted a link to:
https://www.science.org/content/blog-post/mutate-em-all-and-...
which comments the results of this study:
https://www.biorxiv.org/content/10.64898/2026.07.25.740675v1
That study has done in reality what the AlphaGenome Atlas does in fiction, but instead for a human they have done it for one of the simplest viruses.
So they have fuzzed the virus by mutating one by one each position of its DNA.
And various dedicated AI models all made poor predictions of the results of that experiment, which casts doubts about the value of the AlphaGenome predictive map.
A virus is much simpler than a human, but even for that simple virus the effects of most of the mutations could not be predicted. A half of the mutations had harmful effects, and for a half of those it is unknown for now why they were harmful.
For a human the uncertainty about the effects of a mutation will be far greater than for one of the simplest viruses.
> And various dedicated AI models all made poor predictions of the results of that experiment, which casts doubts about the value of the AlphaGenome predictive map
I would not group AlphaGenome into the pile of failed predictions of other models. AlphaGenome deserves to get evaluated based off its own merits.
Yep. Sequence-to-function models are still very limited. AlphaGenome Atlas, despite the flashy branding, is unlikely to provide significant benefit to researchers.
May be that's the reason for the alpha naming. We are waiting for a stable release (just kidding).
Not a word about promoter sequences.
Imagine cellular activity as an industry zone, its not just what you can or can not make, its also 'for what concentrations of chemical species, what transcription rates should be used' so apart from the discrete Mendelian aspects (like what eye color or what have you) there is also a concensus sequence and deviations from consensus. They mention the dataset captures non-coding DNA, which should imply promoter sequences. Will it be possible to query the atlas for joint probabilities of promoter and putative target protein occurence in human genomes?
Personalized medicine could never credibly take off as long as promoter sequences were excised before sequencing!
Eye color is not discrete Mendelian. That's only correct to the first order.
Also Mendelian has little to do with promotor sequences or differential transcription in deviations.
Don't be put off by the box asking for your "affiliation". I wrote "None", clicked submit and it took me straight to the Atlas.
Can this be used with a 23andMe genome to find pathogenic mutations?
23andMe and similar companies don't transcribe your entire genome because that would cost way more than they charge you. They just sample a few tiny sections of it.
A few as in tens of thousands.
23andMe used a custom Illumina Infinium microarray designed around segments of particular interest.
https://www.illumina.com/products/by-brand/infinium.html
fwiw sequencing your entire genome only costs $400 or so from providers like sequencing.com
Thank you for the recommendation! I've been on a lookout for an affordable full sequence and 400$ is very reasonable. Very excited to just have my DNA on my own computer - this feels like the good kind of sci-fi :)
What are their data privacy and security policies? That is an important consideration.
Probably not any 23andMe haven't already told you about. They test a limited set of SNPs, balancing between ones thought useful for genealogy, ones useful for ethnicity estimates and ones thought useful for health-related things (the latter they would like to make their main selling point, the two former are really all commercial DNA services' bread and butter).
It's unlikely that they would luck into testing some unknown SNP which turned out to be relevant for disease.
23andMe tests SNP's (single nucleotides) that are inferred to be significant in protein function/epigenitics.
Those SNP's i believe are testd from primers
so what 23andMe does is specifically on the back of previous research and afaik their data isnt technically clinically significant as most findings need confirmation or more tests.
Not really, no.
Why not? There is no logical reason as to why this would not work, IF it works in the first place, which I don't know. In theory the problem space here is finite, so there is of course a way to predict everything. Whether this is the case right now - who knows; I probably don't think it is currently ready. But eventually it will be. And it should not be in the hands of private companies.
So it sounds like you're coming to this with very little knowledge about biology. I encourage reading up on modern challenges in pathogenic prediction, especially with regards to SNPs: on their own, with the exception of a few diseases, individual SNP predictions are meaningless in terms of actual pathogenicity.
Because 23andme does not sequence your genome, only substring matches linked to specific gene variants.
Alpha Fold has continued to impact the field of protein networks, but I do hear that not every one of the deep learning biology models from Google/Deep Mind and others have made equivalent impact or had as lasting relevance in their respective domains..some have performed more poorly than other available models. I'd love to learn more about this, but this has mostly come from little snips of conversations here and there, in person and online, but I haven't seen anything comprehensive in terms of evaluating their impacts overall
This may not be anything new but it makes using several Google/DeepMind resources a lot less painful.
I'm comfortable programming but others who also do mol bio may be less so or may not recognize when Claude is going off the rails.
People are upvoting this because it has the “Alpha______” prefix. Meanwhile, everyone in the field of genomics knows that AlphaGenome provides essentially zero improvements over the previous SOTA, Borzoi…
> a database that predicts the effects of every possible single nucleotide variant in the human genome. We used the AlphaGenome AI model to pre-calculate the regulatory impact of all 9 billion single-letter genetic changes, resulting in a massive, 1-petabyte dataset.
This is for a database, no? While Borzoi is a model?
> Here, we introduce Borzoi, a model that learns to predict cell-type-specific and tissue-specific RNA-seq coverage from DNA sequence.
https://www.nature.com/articles/s41588-024-02053-6
Any model can be expressed as a database.
Kolmogorov looks at this with a ‘duh’ face
> Meanwhile, everyone in the field of genomics knows that AlphaGenome provides essentially zero improvements over the previous SOTA, Borzoi…
Can you elaborate on this? I'm confused why Google would build something that provides zero improvements over SOTA, Borzoi... as you mention. I'm not familiar with this field, just curious.
Speaking as somebody who has worked within Google Research before: the researchers are under tremendous pressure to publish SOTA and sometimes they juice their results a bit to look competitive when they can't match. This is not uncommon in the field- it's remarkably easy to edit a paper to make yourself look good by omitting information.
One of the most egregious cases of this, in my opinion, is only publishing metrics that cover part of the confusion matrix. “The false-negative rate? That could not possibly matter for a variant effect prediction model; why would we include that in the paper?” Example: AlphaMissense.
The use of an exact quote in an ungrammatical fashion is a bit of a language model smell. I can’t help but be reminded of the purely nonsensical AI interview answers. “It’s a pleasure to meet you, Chick Bongo”
Creating an account 12 minutes ago (from the time of this posting) to comment on how another comment seems to you like a "language model smell" is itself, a language model smell, or a scammer.
Please stop accusing or hinting at others being a language model or bot. Not only is it a dumb waste of time, it's wrong in this instance and you are not only going to continue to be wrong but you have no way to prove or demonstrate that any single post comes from a bot nor the ability to do anything about it if you did in fact believe some comment to be attributed to a bot.
They did benchmark the model and beat the SOTA on every metric though
Is it so bad to have another entrant, especially with the resources Google could bring to bear?
Imagine if the Apple EV had actually happened, you think the EV enthusiasts would roll their eyes like you are?
Are you really taking a holier than thou approach on a google article?
Is this just Google precomputing Alpha genome values - which were already accessible via API and making them available as another API (presumably more broadly)? Or is there actually new information?
That’s my reading. (That this is a cached database of Alpha values)
This has Demis written all over it. There is a great video of him with AlphaFold chatting with the team about releasing some results, and he asked something like “what if we just do them all?”
Very excited to see that happen here.
I don’t think Demis played a big role in this. It was mainly Ziga Avsec who developed Enformer (the first actually decent sequence-to-function model), and then AlphaGenome.
This has nothing to do with AlphaFold at all- not sure if you were implying that. (the scientific contriution is welcome, but it's not particularly significant)
Why are you excited about this?
I saw a really interesting talk by Katie Pollard at ISMB this year about the limitations of variant prediction.
The gist was, can existing variation provide enough context to infer impact of variation? The answer seemed to be no.
Kind of like how frontier LLMs need to ingest larger and large amounts of text to advance. We are going to need to leverage comparative data from other species, and likely tremendous amounts of laboratory mutagenesis experiments to actually make headway on variant prediction. Nature, as it stands, just doesn't have enough human variation.
That's an interesting statement: "The gist was, can existing variation provide enough context to infer impact of variation? The answer seemed to be no."
Is this saying that if we were to sequence every human being on the planet, we'd still be unable to explain some phenotype differences caused by variation simply becase there aren't enough humans/enough variation? Interesting, as that's the first time I've heard that claim, and it would suggest that we spend our time working on mechanistic models of variant to phenotype.
Yes, and genetic variant generally do not act in isolation. We currently focus on the small additive effects of variants because we can with small sample sizes—and 1 million humans is marginal using a GWAS cohort to dive into epistasis. But these interaction effects among variants are critical. Now almost completely deprecated.
I really love Deep Mind, its genuinely focused on using AI to make the world a better place.
Be careful when loving shareholder-driven endeavors; they're a single executive decision away from breaking your heart.
It's "shareholder-linked" (probably not "driven"), but yes, upvoted. Nonetheless, even when with grave faults in the path, Google has managed to give us marvellous, paramount free resources (Google Street, Google Museum...), and would have given more (Google Books - if the negotiation had succeeded). I am grateful.
Google Streetview/maps isn't even the best mapping/'streetview' tool at this point. Filled with ads and bugs, some major bugs have been there for 10+ years.
Now I'm curious. What is the better mapping and streetview tool at this point?
For maps there's quite a few, just anything that doesn't have the amount of ads beats it. For Streetview apple lookaround is a better experience if it's mapped well in your area.
I believe Google has been trying to improve their Streetview in coverage as apple has become a concurrent, but it still has the same bugs people have been complaining about for years (like going backward when pressing forward type of bad). If I remember correctly the bugs have been 'closed' or something before just to include it in the same slightly different way.
As someone that just started working on Google Street View, do you links to any of those bugs? Curious to take a look and see what those problems are
Organic Maps for maps
Why would you be grateful? I don't think any of that should be under private control to begin with.
And which public enterprise has undertaken those projects?
Looking at current hot problems, and exemplary of the situation: which public entities have endeavoured to backup the most important channels and videos from YouTube, the current historic video archive?
Why would I be grateful?! Because somebody did create those resources and made them available! And they did to common good (if anybody refused the expression "for" common good).
They said should be under public control, not is under public control.
Double check.
Be careful when loving government/philanthropy/ngo-driven endeavors; they're a single executive decision away from breaking your heart.
Marriages too
Hopefully you have a bit more impact on the decision making of your marriage shareholders than on public company executives.
Statistically you are very correct.
No AI company is doing that. There is some benefits but that’s not what any AI for profit will ever focus upon
DeepMind objectively is doing that. The world is not so black and white as "evil profit-hungry corporations will never do anything good".
I preemptively acknowledged that in my comment…
If Google does anything good for the world it is entirely accidental, and I am certain they will correct this mistake later.
Market reward is the best proxy for good we have.
What's your alternative? Executive fiat? Mob rule?
Deepmind is objectively good. These endeavors would always cost money. Better private capital than state, no one has to give Google anything. The state is not so generous.
Mob rule, aka Democracy.
Yes, we can just vote ourselves largess from the public purse.
Not what democracy is.
Every single serious theorist of democracy has recognized this.
Democracy is about rule of law and institutions, not simply the mob.
Tyranny of the majority is the paramount concern of all democracies. Nevertheless, institutions can fail.
The idea that democracy always leads to desirable outcomes is evidently false by historical record.
I just don't understand people who think the state is better than Google. Google pretty much works. Google does not have police powers. Google cant surround your car and point guns at your family. Most of the bad things about Google are actually forced upon them by the state. Who could have even a passing knowledge of history and think states are somehow safer than corporations?
Take the United States. Should Donald Trump get a say in how to run Google? Maybe you like him, so how about the guy that comes after him, or the next one after that?
The American mob seems intent on leading their country into the gutter. Good thing (despite what the mob has thrown at it) that the United States still has strong institutions and strong markets. Without Silicon Valley, America would have already lost its place on the world stage.
But please lets hear more indictments from a population who can't even agree with itself on the issues already under the government's purview. Now let's set a precedent for even more state interference in the market, that way when your party (whatever side you are on) chokes the next future election away the next nut you don't agree with has even more power? Is that your plan?
> Should Donald Trump get a say in how to run Google?
Trump does have a say in how to run google, in a very real sense.
> The American mob seems intent on leading their country into the gutter. Good thing (despite what the mob has thrown at it) that the United States still has strong institutions and strong markets. Without Silicon Valley, America would have already lost its place on the world stage.
American elites are more than happy to follow basically whatever Trump wants them to do with the knowledge that they'll get preferential treatment by the law and win all the no-bid contracts he can award. But ask yourself if this would be any different if Trump were head of a large company under a smaller regulatory state. Seems pretty obvious that not much would change.
If by "the mob" you mean, a small cohort of wealthy, powerful, unscrupulous people, those people who make decisions that the rest of us have no say in, then I agree, that is a problem.
> But please lets hear more indictments from a population who can't even agree with itself on the issues already under the government's purview. Now let's set a precedent for even more state interference in the market, that way when your party (whatever side you are on) chokes the next future election away the next nut you don't agree with has even more power? Is that your plan?
In a smaller state, Trump would have continued to be an executive and the shape of his impact on the world might be slightly different, but would be no less oppressive. Ask tenants in NYC in the 1980s if they think a Donald Trump with smaller regulatory oversight who stayed in business would be better or worse for our country.
Do you think our country flips parties every year because our government does a good job of reflecting the will of the people?
You make some excellent points. Let's dig in!
> Trump does have a say in how to run google, in a very real sense.
Indeed, as does every President or government. But surely, you have noticed and take seriously that he has acted with greater executive fiat? That by American standards, while he may not be totally alien, he is not "normal"? The question then is whether we want to accelerate along this trend. What is interesting is many populists, whether on the right or left, seem to answer in the affirmative, confident giving their preferred president increased power is the right trade to make. I simply ask, what happens when the government flips, as it always does? Then maybe you will come to regret your position.
> American elites are more than happy to follow basically whatever Trump wants them to do with the knowledge that they'll get preferential treatment by the law and win all the no-bid contracts he can award. But ask yourself if this would be any different if Trump were head of a large company under a smaller regulatory state. Seems pretty obvious that not much would change.
The first part of this is ahistorical. American elites always line up behind the President. If Kamala had won Zuckerberg would be out front with DEI or w/e. The big companies, banks, Silicon Valley, always shower money and gifts on the incumbent. The government operates in many ways as a big public/private partnership. Again, the danger with Trump is the degree, not necessarily a wholesale difference. Some people use this to say it doesn't matter, makes no difference. I'm not one of those people. I believe at a certain point, tip the degree too far, and a difference appears.
As for the second part, democracy allows the people to correct for their mistake. The elites did not want Trump, and never have. He came to power after descending an escalator, elites on both sides mocked him. Elites wanted to repeat Bush v Clinton. Trump may be wealthy, but he is not an "elite", he is new money, and has basically always been a joke among the upper echelons of power in New York City. The elites are nice to him because they are forced to deal with who the mob elected. To not do so is shareholder suicide. Everyone knows he is insane. They simply cannot say it out loud. And why would they? What's with people who simultaneously think Google has too much power and yet want them to be out in front fighting Trump? It's not their job. Why did the American people elect him? What is Google supposed to do? Ignore the President of the United States? Tell the American people who to vote for?
> If by "the mob" you mean, a small cohort of wealthy, powerful, unscrupulous people, those people who make decisions that the rest of us have no say in, then I agree, that is a problem.
No I mean the ~100,000 average Americans who actually make a difference in electing presidents, in places like Wisconsin, where the state is sufficiently purple and there is a small "swing vote" that is often well under a million people. When are they going to be held responsible for their mistake? Why is Google my enemy? I can stop using Google. Google provides me excellent services at a reasonable price (often $0 - I understand I give them data but who cares, really only the government would care about my individual data, and I know Google simply aggregates it to "subsidize" the numerous free services I get). So, a pretty fair deal, and one I can not enter to if I actually wish. On the other hand, what does someone in Wisconsin give me? I can't stop someone in Wisconsin from voting against my interests. Is Google supposed to be the tyranny here? Extend this same logic to any state on the wrong side of your preferred electoral divide. Like many Americans, I really have no say in Presidential elections, because I live in a populous, blue state. My vote doesn't matter, my electoral votes are already priced in. Meanwhile people in red flyover states get the same number of senators as my state, which has a multiple of their population. This makes corporations my enemy? Did Google sponsor the Connecticut Compromise?
> In a smaller state, Trump would have continued to be an executive and the shape of his impact on the world might be slightly different, but would be no less oppressive. Ask tenants in NYC in the 1980s if they think a Donald Trump with smaller regulatory oversight who stayed in business would be better or worse for our country.
Let's not move the goalposts. I never said I was against all government regulation. If you want to regulate Google, then fine. Make a sensible proposal. What is not serious nor sensible is saying "Google is evil" or "Google should not exist".
> Do you think our country flips parties every year because our government does a good job of reflecting the will of the people?
You need to not think only of the current case relative to your supposed best case, but also a worse case. What is the will of the people? Who decides? How? It seems obvious to me the government being defective is somewhat by design (separation of powers) because the founders knew as many people know now, that America is very large, very divided, and very dangerous. A federal government that doesn't give some of the voters 100% of what they want every four years is a lot better than another civil war. Yes things could be better. They could also be a lot worse.
I tend to think people who say "Google is evil" have a rather limited view of the downside risks that occur when humans congregate. Google does not even begin to scratch the surface of the historical depravities we have visited upon each other.
Have you been to states that predominantly vote against your particular preferred political party? You will find "the corporations" are not your enemy. It's what it always has been, all the time, throughout every historical age. The problem of civilization is other people. You will not get around that by trying to treat corporations as some special case.
Propose sensible regulations. Maybe even a way to reform the electoral college that has a chance of passing. I am interested in these ideas. Innovation can happen in technology and in government. That's why I like this forum. But too often interesting conversations are drowned out by "Google is evil" as if this is useful. I would love to hear more interesting solutions, and less hyperbole. Thanks for the pushback.
> Market reward is the best proxy for good we have.
Only if the goal is profit. All else be damned.
> What's your alternative? Executive fiat? Mob rule?
It is always bewildering to me that people see the world being raped by corporations and their negative externality, and they think that this is the pinnacle that humanity can achieve.
Nothing else can be better, if all your options are "mob rule" or "executive fiat".
I fancy myself a misanthrope, but damn, next to people in this forum I am probably the last humanist walking through the ruins of civilization.
> Deepmind is objectively good.
Google is objectively bad, and the world would be a much nicer place if they ceased to exist.
> It is always bewildering to me that people see the world being raped by corporations and their negative externality, and they think that this is the pinnacle that humanity can achieve.
> I fancy myself a misanthrope, but damn, next to people in this forum I am probably the last humanist walking through the ruins of civilization.
A humanist who uses a term of sexual violence so cavalierly?
If you are in the United States and you want a good civic lesson in state power, you should go to your nearest metropolitan criminal court. Maybe you will have a more accurate view of what government means then. At the very least, you will likely learn to banish such a casual invocation of that word from your lexicon.
> Nothing else can be better, if all your options are "mob rule" or "executive fiat".
What we have now is the middle ground between king and mob. It took centuries and much carnage to achieve. Does the famous Churchill quote really need to keep being repeated? Do you have an alternative proposal? Or maybe you have a sensible proposal that falls within the established democratic bounds? Or perhaps instead by "would be a much nicer place if they ceased to exist" we should simply infer a generalized whining rather than any specific policy prescription?
What's your actual idea, do you have one?
Maybe if you endeavored to communicate without self-satisfied hyperbole you would find we actually have reason to agree with each other.
Executive fiat, given the right executive body, and undergirded by healthy autarkies or other cell-type self-direction, worked well in most locales for most of the years humans have been humans. Certain flashy civilizations, Sumer, Greece, have tended to draw the spotlight incommensurately with their population sizes. But even these functioned much more like what I describe above than anything modern.
I found this one of the more entertaining formulations of "back to the land" style sentiments that I have ever come across here, so the downvotes you are getting aside, I thank you for the humor.
The discussion as I understood it was alternatives to modern capitalism.
Supposedly whilst keeping (something close to) our current levels of welfare and life expectancy.
Solutions that worked for hunter gatherer civilisations 5000 years ago may no longer be viable, since they are incompatible with global supply chains.
I think they have done some quantifiable good.
Alphafold or Earthquake alerts for example spring to mind.
You could argue about comprehensive "free" services (search, Gmail, YouTube, docs, etc) and "free" browsers and "free" operating systems (yeah plural) being net positives in the world but I know that many people view these as negatives as they (rightly or wrongly) view them as vehicles to sell ads or collect data, but I bet you 100 bucks the typical person on the street using their android phone to search Google, watch YouTube, check their Gmail, write a doc, view photos etc doesn't give a shit about that and values these things - google has ended up as a monopoly as these things are actually pretty useful to people and thus valued, even if we in the tech world disagree with the adtech and data aspects. This monopoly was not a fait acompli or pure market abuse - Nokia, MS, flikr, Hotmail, yahoo etc were all incumbent and market dominant but are now irrelevant because the Google products were viewed as better by end users.
I am freaking out. This is a huge moment.
I don't want to drag the discourse away from this achievement, but I hate how this is announced with blatant corporate advertising (our internal model, here are the benchmarks, gpt astra TM yours now for the low low price of £200pcm). I just didn't think Navier-Stokes falling would be sponsored by McDonald's.
Still. I am crying right now. Navier-Stokes is solved.
This post is not about Navier-Stokes. That is in a different thread. Glad you're cry-happy though.
Wait, is NS solved? I thought I was just a singularity thing?
There is an announcement by OpenAI https://news.ycombinator.com/item?id=49613262
So I will know which DNAs to change to become a wolverine! yay!
There are AI labs headed by marketing CEOs that fake metrics, have an utter disregard for humanity and make up "AGI is imminent" propaganda for their IPO, and then there are AI labs headed by actual scientists that do actual science for humanity without constantly trying to put themselves into the spotlight.
Interestingly the labs led by marketing CEOs have much better models (astra is A LOT better than gemini).
Anyone know if you can use an indel VCF file with this?
Cool. But is this actually advanced biology, or just making predictions about biology more accurate? Those aren't the same thing.
I'm convinced AI labs are absolutely hallucinating with their random names. It doesn't even mean anything anymore. Can we go back to numbers?
could this be used with a nebula genomic sequence to find pathogenic sequences?
This Google blog post is a distilled version of a Deep Mind blog post:
https://deepmind.google/blog/alphagenome-atlas-a-predictive-...
They are only announcing a cache. The origin for the cache is not discussed.
In particular, the question of whether to trust the predictions is not addressed. For that, I think the citation is from January:
https://www.nature.com/articles/s41586-025-10014-0
Appreciate the load-bearing info.
hmm gate keeping the database to elitist institutions and private businesses, I'm excited for the future!
All controlled by an adCompany.
That is outright scary. Science is being slurped up here.
unleash gemini 4 stop saving dario
the Google DeepMind PR team can't catch a break, shame it makes no sense to me - if someone's in the field maybe they could explain to the rest of us if this is a big deal, just a PR move, or nah?
Just wait a few months, if it is a big deal that will likely be obvious by then
i like this way of thinking ;)
Does lots of good to my anxiety level :)
What are you even talking about? What do you mean catch a break? They’ve made massive contributions over the last decade.
I think orliesaurus meant the PR team is unable to rest because the rest of Deepmind is so prolific and keep coming out with new things for them to announce
This is exactly what I meant!
I don’t think so. The tone of the comment seems derogatory.
No, it's exactly what @dronebetter said
not my field so I can't judge how useful it is but assuming this data can be used for drug discovery and their ToS limiting to non-commercial use only. Does DeepMind plan on selling this data to pharmaceutical companies?
Individual SNP predictions are not useful for drug discovery. Pharma might license this, but mainly out of fear of missing out.
Probably. I know Google has partnerships with Pharma: https://www.merck.com/news/merck-and-google-cloud-partner-to...
This is just typical enterprise "use AI" infrastructure, not hyper specialized drug research.
My understanding is that they already are, via Isomorphic Labs
It's a bigger deal for diagnostics than for drug development.
Hopefully. Tired, as a shareholder, to see the giveaways that deepmind is doing.
Yeah, screw enabling life changing medical research, won't someone please think of the investors' financial returns??
You can do both. Release it, but ask for money.
Astroturfing, eh? This is verbatim the same comment as another user in this thread. Strange...
PunchTornado was first, fr2029 was second
i made no such comment. reported for harassment.
so you just happen to be posting in a thread where some random user said you made a comment but you made no comments whatsoever
You deleted the verbatim comment after it was pointed out. One was in thread and one was a stand alone comment. I was curious which was the throwaway so clicked into both profiles
Google doing work to uncover the pandoras box of genetics? How long before they shove this under the rug...
An atlas of the human genome from the company whose other atlas still routes me into a lake.