Google DeepMind has released AlphaGenome Atlas, a catalogue of precomputed predictions for the molecular effects of every possible single-nucleotide variant in the human genome. That is roughly 9 billion single-letter changes. The release also introduces the AlphaGenome Variant Impact (AVI) score, a single number that ranks variants by predicted impact, plus per-variant feature attributions and a genome-wide motif collection. The resource ships as a free web portal for academic use, through the AlphaGenome API, and as a skill in Google Antigravity.
Is it deployable? Partially. The Atlas is queryable today for non-commercial research via the portal and API, and commercial access on Google Cloud is listed as “coming soon”. The underlying AlphaGenome model is already available for academic use on GitHub and for commercial use on Model Garden on Google Cloud.
From one model to a genome-wide map
AlphaGenome, released in June 2025, predicts how a DNA variant changes molecular processes such as gene expression and RNA splicing. It has been used widely, but always one variant or one region at a time. The Atlas changes the unit of work. DeepMind team ran AlphaGenome across all 9 billion single-nucleotide variants and stored the outputs, producing a 1-petabyte dataset. This is more than 30 times larger than the AlphaFold Database, which holds over 200 million protein structure predictions.
Testing 9 billion mutations in a lab is not feasible, and running a large model on demand for each candidate variant is slow for genome-scale studies. A lookup table with attached interpretation removes both bottlenecks.
What is inside the Atlas
The Atlas exposes 4 linked resources:
Molecular effect predictions: thousands of predictions per variant, covering multiple aspects of gene regulation across hundreds of human and mouse cell types and tissues.
AVI score: a single impact number per variant. It combines AlphaGenome’s regulatory predictions with AlphaMissense, DeepMind’s model for protein-altering variants, so it works in both coding regions (about 2% of the genome) and non-coding regions (the other 98%).
AVI feature attributions: each score is decomposed into additive contributions from interpretable categories such as chromatin accessibility, splicing, and conservation, so a researcher can see which process a variant is predicted to disrupt.
DNA sequence motifs: a compendium of over 2,500 recurrent short sequences, with genomic locations, including transcription factor binding sites.
DeepMind team reports that the AVI score delivers best-in-class performance across many variant pathogenicity and rare disease benchmarks. The technical report carries the benchmark details.
AlphaGenome Atlas explainer
:root{
–plum:#ffffff; –plum2:#f3f7ff; –line:#cfdcf5;
–lav:#8fc4ff; –pink:#1a5cff; –orange:#0b0b0f; –cream:#0b0b0f; –muted:#4a5568; –mtp:#76B900;
}
*{box-sizing:border-box}
body{margin:0;background:var(–plum);color:var(–cream);font-family:”Helvetica Neue”,Arial,sans-serif;line-height:1.5}
#aga-wrap{max-width:860px;margin:0 auto;padding:22px 18px 16px}
h1,h2{font-family:Georgia,”Times New Roman”,serif;font-weight:400;letter-spacing:.2px;margin:0}
h1{font-size:28px;line-height:1.2}
h2{font-size:19px;margin-bottom:6px}
p{margin:6px 0 0;color:var(–muted);font-size:14px}
.lede{color:var(–cream);font-size:15px;margin-top:8px;max-width:640px}
.panel{background:var(–plum2);border:1px solid var(–line);border-radius:14px;padding:18px;margin-top:16px}
/* strand */
.strand-head{display:flex;justify-content:space-between;align-items:flex-end;gap:12px;flex-wrap:wrap}
.region-key{display:flex;gap:14px;font-size:12px;color:var(–muted)}
.region-key i{display:inline-block;width:10px;height:10px;border-radius:2px;margin-right:5px;vertical-align:-1px}
.strand{display:grid;grid-template-columns:repeat(24,1fr);gap:4px;margin-top:14px}
.base{position:relative;aspect-ratio:1/1.35;border-radius:6px;border:1px solid var(–line);background:#ffffff;color:var(–cream);font:600 14px/1 Georgia,serif;display:flex;align-items:center;justify-content:center;cursor:pointer;transition:transform .15s,background .2s,border-color .2s;padding:0}
.base:hover,.base:focus-visible{transform:translateY(-3px);border-color:var(–lav);outline:none}
.base.coding{background:#e3eeff;border-bottom:3px solid var(–orange)}
.base.noncoding{border-bottom:3px solid var(–lav)}
.base.mut{background:var(–pink);color:#ffffff;border-color:var(–pink);animation:pop .5s ease-out}
.base .pos{position:absolute;bottom:-16px;left:0;right:0;font:10px/1 Arial,sans-serif;color:var(–muted);text-align:center}
@keyframes pop{0%{transform:scale(.6)}60%{transform:scale(1.18)}100%{transform:scale(1)}}
.ripple{position:absolute;inset:-6px;border-radius:10px;border:2px solid var(–pink);opacity:0;pointer-events:none}
.base.mut .ripple{animation:rip .9s ease-out}
@keyframes rip{0%{opacity:.9;transform:scale(.8)}100%{opacity:0;transform:scale(1.9)}}
.readout{display:grid;grid-template-columns:200px 1fr;gap:18px;margin-top:30px;align-items:start}
.avi{text-align:left}
.avi .lab{font-size:12px;color:var(–muted)}
.avi .num{font:400 44px/1 Georgia,serif;color:var(–cream);margin:4px 0 8px;transition:color .3s}
.meter{height:10px;border-radius:6px;background:#e3eeff;overflow:hidden}
.meter b{display:block;height:100%;width:0;border-radius:6px;background:linear-gradient(90deg,var(–lav),var(–pink),#0b0b0f);transition:width .7s cubic-bezier(.2,.8,.2,1)}
.verdict{margin-top:8px;font-size:13px;color:var(–cream)}
.attr .row{display:grid;grid-template-columns:150px 1fr 44px;gap:10px;align-items:center;font-size:13px;margin:7px 0}
.attr .row span:first-child{color:var(–muted)}
.attr .bar{height:9px;border-radius:5px;background:#e3eeff;overflow:hidden}
.attr .bar b{display:block;height:100%;width:0;border-radius:5px;transition:width .6s ease}
.attr .v{text-align:right;color:var(–cream);font-variant-numeric:tabular-nums}
.c-splice b{background:var(–pink)} .c-expr b{background:var(–orange)} .c-access b{background:var(–lav)} .c-cons b{background:#5b7bff} .c-prot b{background:#0b0b0f}
.hint{margin-top:12px;font-size:12px;color:var(–muted)}
.hint em{color:var(–pink);font-style:normal}
/* steps */
.steps{display:grid;grid-template-columns:repeat(4,1fr);gap:8px;margin-top:12px}
.step{background:transparent;border:1px solid var(–line);color:var(–muted);border-radius:10px;padding:10px 8px;cursor:pointer;text-align:left;font:13px/1.3 “Helvetica Neue”,Arial,sans-serif;transition:.2s}
.step small{display:block;font:11px Georgia,serif;color:var(–pink);margin-bottom:4px}
.step[aria-selected=”true”]{background:#e3eeff;border-color:var(–pink);color:var(–cream)}
.stepbody{margin-top:14px;min-height:96px;animation:fade .4s}
@keyframes fade{from{opacity:0;transform:translateY(4px)}to{opacity:1;transform:none}}
.stepbody h3{font-family:Georgia,serif;font-weight:400;font-size:17px;margin:0 0 4px}
.stepbody p{color:var(–cream)}
.flow{display:flex;align-items:center;gap:6px;margin-top:12px;flex-wrap:wrap}
.chip{border:1px solid var(–line);border-radius:999px;padding:4px 10px;font-size:12px;color:var(–muted)}
.chip.on{border-color:var(–pink);color:var(–cream);background:#e3eeff}
.arrow{color:var(–line)}
/* stats */
.stats{display:grid;grid-template-columns:repeat(4,1fr);gap:10px;margin-top:16px}
.stat{background:var(–plum2);border:1px solid var(–line);border-radius:12px;padding:14px 12px}
.stat b{display:block;font:400 26px/1.1 Georgia,serif;color:var(–cream);font-variant-numeric:tabular-nums}
.stat span{display:block;font-size:12px;color:var(–muted);margin-top:4px}
.stat i{display:block;height:3px;width:0;margin-top:10px;border-radius:2px;background:linear-gradient(90deg,var(–lav),var(–pink));transition:width 1.2s ease}
.stat.in i{width:100%}
.foot{display:flex;justify-content:space-between;align-items:center;margin-top:16px;font-size:12px;color:var(–muted);gap:10px;flex-wrap:wrap}
.foot .brand{color:var(–mtp);font-weight:700}
@media (prefers-reduced-motion:reduce){*{animation:none!important;transition:none!important}}
@media (max-width:640px){
h1{font-size:23px}
.strand{grid-template-columns:repeat(12,1fr);gap:3px}
.base{font-size:12px}
.base .pos{display:none}
.readout{grid-template-columns:1fr;gap:12px;margin-top:16px}
.attr .row{grid-template-columns:110px 1fr 40px}
.steps,.stats{grid-template-columns:repeat(2,1fr)}
}
Tap a DNA letter. See what AlphaGenome Atlas does with it.
The Atlas already holds a precomputed prediction for every one of the 9 billion single-letter changes in the human genome. This demo shows what a single lookup returns.
1. Mutate one base
Click any letter to swap it. Coding bases get an AlphaMissense protein term; non-coding bases rely on AlphaGenome alone.
coding (2%)non-coding (98%)
AlphaGenome Variant Impact (AVI)
0.00
No variant selected.
RNA splicing0.00
Gene expression0.00
Chromatin accessibility0.00
Conservation0.00
Protein (AlphaMissense)0.00
Illustrative numbers. The bars mimic how the Atlas splits one AVI score into additive feature attributions. Real values come from the Atlas portal, not this widget.
2. How the Atlas is built
Step 1Precompute effects
Step 2Collapse to AVI
Step 3Attribute the score
Step 4Map the motifs
9B variants→
AlphaGenome→
thousands of molecular effects→
AVI score→
attributions + 2,500+ motifs
0single-nucleotide variants scored
0dataset size, 30x the AlphaFold Database
0more non-coding associations found in 54,000+ UK Biobank genomes
0recurrent DNA motifs catalogued
Source: Google DeepMind, AlphaGenome Atlas announcement, Sept 8, 2026. Not for clinical use.Built by Marktechpost
(function(){
var seq=”ATGCGTACCGATTGCAGGCTTACG”.split(“”);
var coding=[0,1,2,3,4,5,6,7,8,9,10,11,12]; // first 13 bases shown as an exon
var strand=document.getElementById(“strand”);
var letters=[“A”,”C”,”G”,”T”];
var current=-1;
seq.forEach(function(b,i){
var el=document.createElement(“button”);
el.className=”base “+(coding.indexOf(i)>-1?”coding”:”noncoding”);
el.type=”button”; el.textContent=b; el.setAttribute(“aria-label”,”position “+(i+1)+”, base “+b);
var r=document.createElement(“span”); r.className=”ripple”; el.appendChild(r);
var p=document.createElement(“span”); p.className=”pos”; p.textContent=i+1; el.appendChild(p);
el.addEventListener(“click”,function(){mutate(i,el);});
strand.appendChild(el);
});
function hash(i,l){var h=(i*2654435761+l.charCodeAt(0)*40503)%1000;return h/1000;}
function mutate(i,el){
var old=el.firstChild.nodeValue;
var nxt=letters[(letters.indexOf(old)+1)%4];
el.firstChild.nodeValue=nxt;
if(current>-1&¤t!==i){var prev=strand.children[current];prev.classList.remove(“mut”);}
el.classList.remove(“mut”); void el.offsetWidth; el.classList.add(“mut”);
current=i;
var isCoding=coding.indexOf(i)>-1;
var s=hash(i,nxt);
var attrs;
if(isCoding){
attrs={splice:s*0.25,expr:s*0.12,access:s*0.05,cons:0.10+s*0.15,prot:0.15+s*0.30};
}else{
attrs={splice:s*0.18,expr:0.05+s*0.30,access:0.05+s*0.25,cons:s*0.12,prot:0};
}
var total=0;for(var k in attrs)total+=attrs[k]; total=Math.min(total,0.99);
document.getElementById(“aviNum”).textContent=total.toFixed(2);
document.getElementById(“aviBar”).style.width=(total*100)+”%”;
var v=document.getElementById(“verdict”);
var region=isCoding?”coding”:”non-coding”;
var top=Object.keys(attrs).sort(function(a,b){return attrs[b]-attrs[a];})[0];
var names={splice:”RNA splicing”,expr:”gene expression”,access:”chromatin accessibility”,cons:”conservation”,prot:”protein impact”};
v.innerHTML=”Position “+(i+1)+” (“+region+”): “+old+”→”+nxt+”. Largest predicted contribution: “+names[top]+”.”;
var rows=document.querySelectorAll(“#attr .row”);
var order=[“splice”,”expr”,”access”,”cons”,”prot”];
rows.forEach(function(r,n){
var val=attrs[order[n]];
r.querySelector(“b”).style.width=(val*100/0.5)+”%”;
r.querySelector(“.v”).textContent=val.toFixed(2);
});
if(!isCoding){rows[4].style.opacity=.4;}else{rows[4].style.opacity=1;}
notify();
}
var stepText=[
[“Precompute, once, for every variant”,”AlphaGenome is run across all 9 billion single-nucleotide variants in the human genome, producing thousands of molecular effect predictions per variant across hundreds of human and mouse cell types and tissues. The result is a 1-petabyte dataset.”],
[“Collapse thousands of predictions into one number”,”The AlphaGenome Variant Impact score fuses AlphaGenome’s regulatory predictions with AlphaMissense’s protein-impact predictions. One score per variant, valid for both coding and non-coding DNA, so researchers can rank candidates fast.”],
[“Explain the number”,”Each AVI score is decomposed into additive feature attributions across interpretable categories such as chromatin accessibility, splicing, and conservation. That is how the DNM1 variant was traced to an incorrect splice site.”],
[“Find the words of the genome”,”A compendium of over 2,500 recurrent DNA motifs, with their locations, links variants to the short sequences they disrupt, including transcription factor binding sites.”]
];
var steps=document.querySelectorAll(“#steps .step”);
var chips=document.querySelectorAll(“#flow .chip”);
function showStep(n){
steps.forEach(function(s,i){s.setAttribute(“aria-selected”,i===n?”true”:”false”);});
var b=document.getElementById(“stepbody”);
b.innerHTML=””+stepText[n][0]+””+stepText[n][1]+””;
b.style.animation=”none”; void b.offsetWidth; b.style.animation=””;
var lit=[[0,1,2],[3],[4],[4]][n];
chips.forEach(function(c,i){c.classList.toggle(“on”,lit.indexOf(i)>-1);});
notify();
}
steps.forEach(function(s){s.addEventListener(“click”,function(){showStep(+s.dataset.i);});});
showStep(0);
// counters
function count(el){
var target=+el.dataset.n, suf=el.dataset.suf, t0=null, dur=1300;
function frame(t){ if(!t0)t0=t; var p=Math.min((t-t0)/dur,1); p=1-Math.pow(1-p,3);
var val=Math.round(target*p); el.textContent=(val>=1000?val.toLocaleString():val)+suf;
if(p