Skip to main content
London Tutorsby KnowledgeDottCom

14 November 2026 · London Tutors

AlphaGenome 2: The AI That Reads DNA One Cell at a Time

Google DeepMind has announced AlphaGenome 2, an AI model that predicts how a DNA change affects one specific cell type at a time — and finds nearly twice as many gene variants that change how genes work.

AlphaGenome 2: The AI That Reads DNA One Cell at a Time

Every cell in your body carries the same DNA — the same roughly three billion letters of genetic instruction. Yet a heart muscle cell and an immune cell fighting an infection behave completely differently. The difference is which genes each cell switches on or off. The hard question in genetics is this: when a single letter of DNA changes, does it matter — and in which cell?

On 11 November 2026, Google DeepMind announced AlphaGenome 2, a new AI model built to answer exactly that question, cell by cell. It is the follow-up to the original AlphaGenome, launched last year, which predicted how DNA variants affect broad tissues. The new version zooms in much further: give it a stretch of DNA up to a million letters long, plus one chosen cell type or cell state, and it predicts how that cell reads the DNA — which genes switch on, how the DNA is packaged, and how chemical marks on the DNA change.

Why "which cell?" changes everything

A DNA variant can be harmless in one cell type and disruptive in another. The old model averaged its predictions across whole tissues, so a variant that mattered only in one rare cell type could be missed entirely. AlphaGenome 2 is what DeepMind calls a conditional model: it takes the cell type as an input at every stage, so the same stretch of DNA is read differently depending on the cell.

It was trained on human and mouse data including more than 107 million individual cell profiles across 68 tissues — a huge library of how real cells actually behave. On a test using TenK10K, a map of 28 primary immune cell types, it identified nearly twice as many variants that genuinely change gene activity as the previous model: 30% against 16%. It also got the direction of the effect right far more often — 43% of the time against 23% — meaning it could say not just "this variant matters" but whether it turns a gene up or down.

A real disease clue, found by the computer

The most striking demonstration came from inflammatory bowel disease. A recent landmark study showed that one variant, sitting in a distant regulatory switch, drives the disease by overactivating a gene called ETS2 in gut immune cells called macrophages. AlphaGenome 2 reproduced that finding on its own: it singled out the same variant and showed it boosts ETS2 in gut macrophages while leaving other cell types and genes largely untouched.

The model can also work with cell states it has never seen before. Given single-cell data from a COVID-19 mRNA vaccine study, it could tell vaccine-activated immune T cells apart from resting ones — and it spotted that a variant linked to severe COVID-19 disrupts gene activity more strongly when cells are in that activated state, an effect that would be missed in resting cells.

What it is — and is not — for

DeepMind is clear about the limits. All the performance figures are the company's own, from its own tests; the technical paper is a preprint, not yet peer-reviewed, and outside labs will need time to test the model themselves. And the model is for research use only: it has not been validated or approved for clinical or diagnostic use, so it cannot diagnose anyone or guide treatment. It is available now for non-commercial research through the AlphaGenome API, with commercial research in private preview.

For students, though, the idea behind it is wonderfully simple and worth understanding: the same instruction book can be read in many different ways, and who is reading matters as much as what is written. That is true of DNA — and, come to think of it, of textbooks too. If your child is studying biology and wants genetics to feel like a living subject rather than a list of terms, find a biology tutor who can bring ideas like this to life, or post a tutoring request and compare tutors for your child.

Book your first lesson in biology and turn headline science into understanding your child can use: Book your first lesson

Sources: AlphaGenome 2: Deciphering how cells read DNA to better understand disease, Google DeepMind, 11 November 2026; AlphaGenome 2 technical paper, Google DeepMind, 11 November 2026

Curious about biology? Post a request on London Tutors and compare verified biology tutors — online, in person or in a group class.

Ready to find the right tutor?

Browse verified tutors, compare profiles, and book a trial lesson — free to sign up.

Find a tutor — sign up free

Related guides