DeepMind has released AlphaGenome Atlas, a free tool that lets drug discovery teams rank genetic variants across the entire human genome without needing expensive computing infrastructure. The database precomputes predictions for all 9 billion possible single-nucleotide variants (SNVs) in the human genome, plus over 100 million insertions and deletions, making variant prioritization accessible to academic labs and small biotechs that previously lacked the resources to run these analyses themselves.

Why Does This Matter for Finding New Drug Targets?

Identifying which genetic variant actually causes a disease is the foundation of modern drug discovery. The challenge is that 98 percent of the human genome does not encode proteins; instead, it regulates gene activity. Running computational models across every candidate variant in those noncoding regions demands graphics processing unit (GPU) infrastructure that most academic groups and small biotechs cannot afford to access on demand. AlphaGenome Atlas solves that bottleneck by precomputing all the hard work upfront and serving results through a simple web portal, free for non-commercial research.

The database is massive: it stores approximately 1 petabyte of data, more than 30 times the size of the AlphaFold Database. For each variant, it contains an average of 27,000 experiment-specific predictions spanning hundreds of human and mouse cell types and tissues.

How Does the Tool Actually Work for Researchers?

The Atlas provides several ways to access variant predictions. The centerpiece is the Variant Impact (AVI) score, which condenses predictions from multiple sources into a single number that ranks how disruptive a variant is likely to be. The score follows a PHRED-like scale: a score of 10 puts a variant in the top 10 percent of predicted impact, while a score of 30 reaches the top 0.1 percent. Each AVI score also includes feature attributions that break down which biological process is most disrupted: RNA splicing, gene expression, chromatin accessibility, or protein function.

Researchers can access the Atlas through multiple channels:

• Web Portal: A no-code lookup interface that requires no programming experience or computing resources.
• AlphaGenome API: For teams that want to integrate predictions into their own analysis pipelines.
• Google Antigravity Integration: Available as a skill within Google's Antigravity platform for cloud-based workflows.

How Much Better Is This Than Existing Tools?

The AVI score significantly outperforms the most widely used genome-wide scoring tool, Combined Annotation Dependent Depletion (CADD). When tested on solved rare disease cases from the Genomics Research to Elucidate the Genetics of Rare diseases (GREGoR) Consortium, a U.S. National Institutes of Health program, the AVI score placed the known causal variant among the top 50 candidates in 29.5 percent of cases. CADD reached only 12.5 percent on the same benchmark, a roughly 2.4-fold difference that matters at the front end of any target identification workflow.

Real-world validation is already underway. At the University of Exeter, statistical geneticist Gareth Hawkes applied Atlas predictions to 54,000 whole genomes from the UK Biobank, grouping rare variants by their predicted molecular effect to surface noncoding associations with blood protein levels. The analysis uncovered 22 percent more noncoding genetic associations than prior methods found in the same data.

"At the Broad Institute of MIT and Harvard, Laura Covill and Anne O'Donnell-Luria applied AVI scores to unsolved rare disease cases and identified a deep intronic variant in DNM1, a gene associated with epileptic encephalopathy. The model flagged the variant as creating an aberrant splice site; experimental validation later supported that call," according to the research.

Research from Broad Institute of MIT and Harvard

What Are the Known Limitations?

DeepMind has been transparent about where AlphaGenome Atlas falls short. The tool has not been validated for clinical use; the accompanying preprint positions it as one input within a broader diagnostic evidence chain, not a standalone basis for clinical decisions. Training data gaps include missing cell types and non-polyadenylated RNAs, and tissue labels reflect broad ontological categories rather than the specific cellular states where regulatory elements can behave differently.

The gaps are most consequential in oncology. AlphaGenome trained on germline reference sequences from general population data, so tumor-specific regulatory contexts shaped by somatic variation and clonal evolution fall outside its scope. Near the TAL1 oncogene in T-cell leukemia, predictions diverged from observed RNA and protein patterns because a short oncogenic isoform was absent from the training annotation.

What Comes Next for Adoption in Pharma?

DeepMind makes AlphaGenome Atlas free for non-commercial research today and plans commercial access through Google Cloud, which will determine how widely the tool spreads through pharmaceutical companies. The Atlas research remains a preprint pending peer review, and DeepMind describes the current release as a first-generation baseline with broader cell-type coverage targeted for future iterations. Used as a prioritization signal feeding into experimental follow-up rather than a standalone endpoint, AVI scores already represent a meaningful step forward for labs that previously had no practical route to genome-wide noncoding variant ranking.

The broader context matters here: while AI has made dramatic progress in theoretical science and mathematical proofs, the translation into experimental biology has been slower. A recent study from Google, Google DeepMind, and MIT found that 44 percent of scientists surveyed said their main research bottleneck had shifted downstream, toward later stages such as physical experimentation and data collection. Forty-one percent said their backlog of untested hypotheses had grown. AlphaGenome Atlas addresses one piece of that puzzle by removing the computational barrier to hypothesis generation, but the real work of validating those hypotheses in the lab still falls to researchers and their teams.