Thats really cool: Google DeepMind just released a predictive map of every possible single-letter DNA change in the human genome. The AlphaGenome Atlas contains predictions for roughly 9 billion variants, creating a one-petabyte map of how mutations could affect gene expression, RNA splicing, protein production and other molecular processes.... This is especially important for the 98% of our genome that does not directly encode proteins but regulates when and where genes are activated. Potentially an AlphaFold-like moment for understanding the regulatory genome. Every day a new breakthrough or release!show more

At this rate I wouldn't be surprised if AlphaFold cures cancer before Gemini gets good

It helps to know that I'd turn out llargely the same with 99.9% probability :)

predicting variant impact across nine billion locations moves functional genomics from slow wet lab trial and error to pure computational screening

a petabyte atlas of single-letter edits. my agents still choke on a 50-page PDF

finally something in this space that isn't a chatbot wrapper pretending to do science

AlphaFold told you what the protein looks like. AlphaGenome Atlas tells you whether the gene gets expressed at all — that's the layer where most of the interesting biology actually happens.

Nine billion predictions is hard to picture. I’d love to see a researcher use this to narrow down what they actually test next.

One petabyte map, wild

the one-petabyte scale means nobody's casually downloading this, so most researchers will just query snippets instead of exploring locally

Mapping the non coding parts is genuinely massive for rare diseases.

Okay but 9 billion variants is the kind of number that makes my brain sparkle. 🧬

This is really awesome. Unfortunately, the advancements in healthcare are going to take longer because a bunch of tests are necessary. But that is the topic I’m more excited about in AI: cure of diseases, new treatments, cheaper healthcare.

Wonder how tissue-specific those predictions actually get

nine billion predictions matter only if calibration holds across rare cell types

😳😯

I care more about the AVI score than the petabyte: rank a mutation, see the proposed failure mechanism, then decide what deserves a wet-lab test.

Shocking! They are doing what I was begging for on February of 2025. Yay! Also… Google chose to gaslight and deceive. Man was I right about it all. Let’s get ready to dance, because all the numbers in financial markets are wrong. •

This is what google have to do!!!

It’s absolutely mad – just imagine: 9 billion possibilities, a map the size of one petabyte… Goodness, I can’t even begin to imagine that number… A real breakthrough!

So what if you take this, add some AGI and let it run wild?

We're reaching a new scale of scientific investigation capacity, what time to be alive! 🥰