Biology
10 labs · 15 models and products Theme: Companies · Generated from this page's data. Click a name to open its panel on the Labs page. Records resting on one source are tagged "one source".AI in biology trains models on DNA, protein and cell data so they can predict what genes and proteins do and which changes cause disease. EMBL-EBI, Europe's bioinformatics institute, runs with Google DeepMind the free AlphaFold database of over 200 million predicted protein structures. Companies on this tile include Basecamp Research, which collects genetic data from nature to train its EDEN models; InstaDeep, owned by BioNTech; Shift Bioscience, which looks for genes that could make ageing cells young again; and Cambridge's Houdini Bio and Tolemy Bio (per the site).
Context
- The AlphaFold Protein Structure Database, run by Google DeepMind with EMBL-EBI, holds over 200 million predicted protein structures. accessed 31 October 2026 alphafold.ebi.ac.uk/about
we’ve released over 200 million protein structure predictions by AlphaFold
- Google DeepMind's AlphaMissense, published in Science, labelled 89% of all 71 million possible single-letter protein changes as likely harmful or harmless. 19 September 2023 deepmind.google/blog/a-catalogue-of-genetic-muta
we show it categorised 89% of all 71 million possible missense variants
- Google DeepMind's AlphaGenome reads DNA stretches up to a million letters long; the research appeared in Nature in January 2026. 25 June 2025 (updated January 2026) deepmind.google/blog/alphagenome-ai-for-better-u
takes a long DNA sequence as input — up to 1 million letters, also known as base-pairs
- A government survey of the UK engineering biology community found about seven in ten respondents use AI tools. 29 January 2026 gov.uk/government/publications/a-survey-into-the
About 7 in 10 (69%) of all respondents use AI tools
- The same survey found trust in AI outputs was the barrier respondents cited most. 29 January 2026 gov.uk/government/publications/a-survey-into-the
Trust in outputs is the most cited barrier to AI adoption by all respondents
Key papers
- Highly accurate protein structure prediction with AlphaFold Nature · 15 July 2021 (online)
Labs Round 8 (31 Oct 2026): introduction and facts as the researchers wrote them; "(per the site)" marks a sentence resting on this site's own description of that lab.
| Record | What it is | Place | Links | Confidence |
|---|---|---|---|---|
| aiPGI (Basecamp Research)Chips, models and products | Basecamp Research's AI gene-insertion platform, powered by its EDEN models, for designing gene-editing enzymes. announcedKey facts Powered by EDEN; active insertion proteins for 100% of tested disease-relevant sites; over 10,000 disease-related genome locations; CAR-T cells over 90% tumour-cell clearance in lab assays (2026-01-12)Made by Basecamp Research · 2026-01-12 | Clerkenwell Workshops, 27 Clerkenwell Close, London (EC1R 0AT) (via Basecamp Research) | prnewswire.com/news-releases/basecamp-research-l | one source |
| AlphaFold Protein Structure DatabaseChips, models and products | A free public database of AI-predicted protein structures, made by Google DeepMind's AlphaFold and run with EMBL-EBI.Key facts Over 200 million protein structure predictions released (accessed 2026-10-30). 16 Mar 2026: 1.7 million high-confidence homodimer predictions added, about 18 million lower-confidence ones for bulk download, 30 million complexes calculated, with partners EMBL-EBI, Google DeepMind, NVIDIA and Seoul National University. 19 May 2026: almost 80,000 high-confidence heterodimer predictions added, 8.1 million lower-confidence ones for bulk download. Over 3.4 million users from 190 countries.Made by EMBL-EBI AI and machine learning · 2026-05-19Key facts and news (1)News
Labs Round 8 (31 Oct 2026), from the researchers' rows; one-source items are marked. | — (Hinxton named in the task brief; not stated on pages opened) (via EMBL-EBI AI and machine learning) | alphafold.ebi.ac.uk/about · ebi.ac.uk/about/news/technology-and-innovation/f · deepmind.google/blog/alphafold-five-years-of-imp | lab's own pages |
| AlphaGenomeChips, models and products | An AI model that predicts how single DNA changes affect the processes that switch genes on and off. released (API preview for non-commercial research)Key facts Theme: Biology. Announced 25 Jun 2025; API preview for non-commercial research, model release planned; Nature paper January 2026. Figures (25 Jun 2025): reads up to 1 million DNA letters; beat best external models on 22 of 24 single-sequence evaluations and matched or beat them on 24 of 26 variant-effect evaluations. Builds on Enformer; complements AlphaMissense. Follow-on: AlphaGenome Atlas (8 Sep 2026).Made by Google DeepMind · 2025-06-25Read the paper | London (EC2M 4AG) (via Google DeepMind) | deepmind.google/blog/alphagenome-ai-for-better-u · deepmind.google/blog/page/6/ | one source |
| AlphaGenome AtlasChips, models and products | A free catalogue predicting the molecular effects of about 9 billion possible single-letter changes in the human genome. released (free portal for academic research; API)Key facts Theme: Biology. Published 8 Sep 2026; free web portal for non-commercial academic use, commercial use on Google Cloud planned; AlphaGenome API and Google Antigravity skill. Figures (8 Sep 2026): 9 billion single-nucleotide variants; 1-petabyte dataset; "more than 30 times larger than the AlphaFold Database"; tested with over 54,000 UK Biobank participants.Made by Google DeepMind · 2026-09-08 | London (EC2M 4AG) (via Google DeepMind) | deepmind.google/blog/alphagenome-atlas-a-predict · deepmind.google/blog/rss.xml | lab's own pages |
| AlphaMissenseChips, models and products | An AI model that estimates whether single-letter changes in human proteins are likely to cause disease. released (open catalogue CC BY 4.0; open code)Key facts Theme: Biology. Science paper 19 Sep 2023. Figures (19 Sep 2023): 71 million possible missense variants scored; 89% classified (32% likely pathogenic, 57% likely benign). Catalogue free for research and commercial use, CC BY 4.0 from 13 Mar 2024; model code open source. Builds on AlphaFold.Made by Google DeepMind · 2023-09-19 | London (EC2M 4AG) (via Google DeepMind) | deepmind.google/blog/a-catalogue-of-genetic-muta · deepmind.google/blog/page/9/ | one source |
| Basecamp ResearchCompanies | Collects genetic data from nature and trains biological foundation models, called EDEN, to design gene therapies, enzymes and peptides.Profile
Labs Round 7 deep profile (30 Oct 2026), as the researchers wrote it; confidence: single_source. Sources Models and products (4)
Selected papers (7)
Key facts and news (3)
Labs Round 8 (31 Oct 2026), from the researchers' rows; one-source items are marked. | Clerkenwell Workshops, 27 Clerkenwell Close, London (EC1R 0AT) | Companies House · basecamp-research.com/news-research/basecamp-res · basecamp-research.com/about · find-and-update.company-information.service.gov. · blogs.nvidia.com/blog/uk-ecosystem-deployment/ | verified |
| CellVoyantCompanies · also on Start-ups | Imaging and AI that predict how living stem cells will behave, to help make cell therapies.Why it is here: Bristol spin-out (2021) whose AI predicts the behaviour of live stem cells from imaging | Bristol (BS8 1HP) | Companies House · uktech.news/tech-hubs/south-west/bristol/bristol · find-and-update.company-information.service.gov. · find-and-update.company-information.service.gov. | verified |
| ConcrCompanies | Cambridge University spin-out that uses astrophysics-derived Bayesian AI "digital twins" to predict how cancer patients respond to treatment.Key facts £1.94m seed 2023 co-led by Cambridge Enterprise Ventures and R42 Group; £0.8m Innovate UK grant for clinical validation; first patients recruited to a breast-cancer observational trial July 2024; 2026-09-03 pancreatic-cancer study in Frontiers in Artificial Intelligence: AUC 72.3% vs 44.8% for conventional biomarkersModels and products (1)
| London (described as "London-based" in trade; Cambridge Enterprise portfolio) | Companies House · enterprise.cam.ac.uk/?p=33665 · precisionmedicineonline.com/precision-oncology/c | verified |
| EDEN (family of biological foundation models)Chips, models and products | Basecamp Research's 28-billion-parameter DNA foundation model family for designing programmable therapies, including gene-insertion enzymes. released (two capabilities live; chat connector in Claude)Key facts Preprint 12 Jan 2026 (bioRxiv, "Designing AI-programmable therapeutics with the EDEN family of foundation models"). Lab model page (read 30 Oct 2026): 28 billion parameters; 9.7 trillion nucleotide training tokens; about 1.95×10^24 FLOPs of compute; two capabilities "live now" for researchers worldwide via a chat interface, with an MCP connector and a Claude directory listing. No open weights stated.Made by Basecamp Research · 2026-01-12 | Clerkenwell Workshops, 27 Clerkenwell Close, London (EC1R 0AT) (via Basecamp Research) | basecamp-research.com/ai-models · basecamp-research.com/news-research | lab's own pages |
| EMBL-EBI AI and machine learningNetworks, bodies, events and media | Europe's bioinformatics institute prepares AI-ready biological data, builds AI tools and runs the AlphaFold Database with Google DeepMind.Key facts AI work described: curating and enriching data for AI, research teams building AI methods, AI training courses (accessed 2026-10-30). Tools and news (titles and dates from its AI news list): BioChatter open-source LLM framework (22 Jan 2025); "InterPro 105.0: AI for protein classification" (28 Apr 2025); "AI model forecasts disease risk decades in advance" (17 Sep 2025); "BioAIrepo: EMBL-EBI's hub for life science AI models" (26 Feb 2026). Runs the AlphaFold Database (BIZ-LAB-6010). Interim Director: Jo McEntyre (16 Mar 2026). Headcount not published.Profile
Labs Round 8 deep profile (31 Oct 2026), as the researchers wrote it; confidence: one source (EMBL-EBI pages). Sources Models and products (1)
| — (Hinxton named in the task brief; not stated on pages opened) | ebi.ac.uk/about/our-impact/ai-and-machine-learni · ebi.ac.uk/about/news/tag/artificial-intelligence · ebi.ac.uk/about/news/technology-and-innovation/f | one source |
| EnformerChips, models and products | Model that predicts how DNA sequence controls gene activity, reading long stretches of DNA at once. released (open model; paper)Key facts Announced 4 Oct 2021; reads 200,000 base pairs of input DNA (more than 5x previous methods); predicts over 5,000 genomic tracks; model and variant predictions openly available; paper in Nature Methods (s41592-021-01252-x); access: open modelMade by Google DeepMind · 2021-10-04 | London | deepmind.google/blog/predicting-gene-expression- | one source |
| FarrSight (FarrSight-Twin) (Concr)Chips, models and products | Concr's Bayesian AI model that builds a "digital twin" of a patient's tumour to predict treatment response. research use onlyKey facts Built on a proprietary Bayesian foundation model drawing on 25 billion data points (2026-09-03); research use only; pancreatic-cancer AUC 72.3% vs 44.8%; used in ICR / Royal Marsden / Durham AI-VISION breast-cancer studyMade by Concr · 2026-09-03 | London (described as "London-based" in trade; Cambridge Enterprise portfolio) (via Concr) | enterprise.cam.ac.uk/?p=39403 · enterprise.cam.ac.uk/?p=33665 | lab's own pages |
| HiFi-NNChips, models and products | A Basecamp Research model for labelling the functions of unknown microbial proteins ("microbial dark matter"). released (paper)Key facts Research item dated 20 Jun 2025 on the lab's news page (journal article in iScience per the Elsevier link; not opened).Made by Basecamp Research · 2025-06-20 | Clerkenwell Workshops, 27 Clerkenwell Close, London (EC1R 0AT) (via Basecamp Research) | basecamp-research.com/news-research | one source |
| Houdini BioCompanies | University of Cambridge spin-out using machine learning to redesign therapeutic DNA so gene and cell therapies stay active longer at lower doses.Key facts Nearly £1.5m pre-seed incl. non-dilutive grants (2026-06-22, oversubscribed); reports more than tenfold higher gene expression than competitors' methods in preclinical dataProfile
Labs Round 8 deep profile (31 Oct 2026), as the researchers wrote it; confidence: verified (funding: Cambridge Enterprise, an investor's parent, plus The Jewish Chronicle); founding year and HQ one source. Sources Key facts and news (1)
Labs Round 8 (31 Oct 2026), from the researchers' rows; one-source items are marked. | Cambridge | uktech.news/medtech/houdini-bio-secures-1-5m-to- · uktech.news/funding/uk-tech-funding-roundup-part · enterprise.cam.ac.uk/?p=38610 · thejc.com/news/uk/houdini-british-jewish-start-a | one source |
| InstaDeepCompanies · also on Start-ups | Builds AI decision-making software for companies, including protein design and circuit-board design tools; owned by BioNTech. acquiredProfile
Labs Round 8 deep profile (31 Oct 2026), as the researchers wrote it; confidence: verified (ownership: company plus trade); models one source. Sources Models and products (5)
Selected papers (5)
Key facts and news (6)
News
Labs Round 8 (31 Oct 2026), from the researchers' rows; one-source items are marked. | London (NW1 3BF ) | Companies House · uktech.news/ai/instadeep-funding-series-b-202201 · find-and-update.company-information.service.gov. · find-and-update.company-information.service.gov. · find-and-update.company-information.service.gov. | verified |
| InstaNovo and InstaNovo-P (InstaDeep)Chips, models and products | InstaDeep transformer models that read protein fragments straight from mass-spectrometry data, including phosphorylated proteins. releasedKey facts InstaNovo trained on 63 million labelled spectra (2025-10-13); InstaNovo-P is a phosphorylation-tuned version (bioRxiv 10.1101/2025.05.14.654049v1; page dated 2026-07-09) with academic co-authors at DTU, Ghent, VIB, Manchester and TU DelftMade by InstaDeep · 2026-07-09 | London (NW1 3BF ) (via InstaDeep) | instadeep.com/2025/10/ai-day-2025-powering-biolo · instadeep.com/research/paper/instanovo-p-a-de-no | lab's own pages |
| MOJO (InstaDeep)Chips, models and products | InstaDeep foundation model that learns from tumour gene activity and DNA methylation together to classify cancers. released (checkpoints on Hugging Face)Key facts Presented at ICML 2026; weighted F1 0.952 for cancer-type classification vs 0.943 for BulkRNABert; survival C-index 0.771 vs 0.749; paper "Bimodal masked language modeling for bulk RNA-seq and DNA methylation representation learning"Made by InstaDeep · 2026-07-08 | London (NW1 3BF ) (via InstaDeep) | instadeep.com/2026/07/mojo-multi-omics-ai-cancer | one source |
| Nucleome TherapeuticsCompanies · also on Start-ups | Oxford biotech using machine learning on the non-coding genome to find new medicines.Why it is here: Oxford spin-out using machine learning to find disease-linked variants in the non-coding ("dark") genome | Oxford (OX2 0HA) | Companies House · uktech.news/ai/nucleome-therapeutics-series-a-20 · find-and-update.company-information.service.gov. · find-and-update.company-information.service.gov. | verified |
| Nucleotide Transformer v3 (NTv3) (InstaDeep)Chips, models and products | InstaDeep's DNA foundation model family that reads, annotates and helps design genome sequences across many species. released (open checkpoints)Key facts Released 2026-02-18; sizes 8M to 650M parameters; pre-training on "more than eight trillion nucleotides"; post-training on ~16,000 functional tracks from 24 species; full model family on Hugging Face; bioRxiv preprint 2025 (DOI 10.64898/2025.12.22.695963v1); InstaDeep AI Day (2025-10-13) gave "15 trillion tokens spanning over 150,000 species"Made by InstaDeep · 2026-02-18 | London (NW1 3BF ) (via InstaDeep) | instadeep.com/2026/02/modelling-the-genome-with- · instadeep.com/2025/10/ai-day-2025-powering-biolo | lab's own pages |
| Ochre BioCompanies | Oxford liver-disease biotech that combines machine learning with large human liver datasets to find RNA drug targets.Key facts HQ Oxford; labs in New York (Alexandria Center, relocated 2025-07-21) and Taipei; partnerships with Boehringer Ingelheim (2024-04-22) and GSK (AI pact 2024-06-12; research phase complete 2025-11-18); proprietary OBELiX software; says it has "invested over $20m" in human liver datasets; Clinical Translation Board announced 2026-10-07 | Oxford | Companies House · ochre-bio.com/2025/11/18/ochre-bio-completes-hum · ochre-bio.com/ | one source |
| SegmentNT (InstaDeep)Chips, models and products | InstaDeep model built on Nucleotide Transformer that marks out genes and control regions in DNA, letter by letter. releasedKey facts Locates 14 classes of genomic elements in sequences up to 50kb; published in Nature Methods ("Annotating the genome at single-nucleotide resolution with DNA foundation models"); splice-site MCC 0.75/0.76 vs SpliceAI 0.67/0.59; on GitHub and Hugging FaceMade by InstaDeep · 2025-10-29 | London (NW1 3BF ) (via InstaDeep) | instadeep.com/2025/10/genome-annotation-with-seg | one source |
| Shift BioscienceCompanies · also on Start-ups | Cambridge company using generative AI to find genes that could make ageing cells young again, as a route to treating age-related disease.Profile
Labs Round 8 deep profile (31 Oct 2026), as the researchers wrote it; confidence: one source. Sources Models and products (1)
Selected papers (1)
Key facts and news (5)
News
Labs Round 8 (31 Oct 2026), from the researchers' rows; one-source items are marked. | Cambridge | Companies House · uktech.news/ai/shift-bioscience-gen-ai-reverse-a · find-and-update.company-information.service.gov. · find-and-update.company-information.service.gov. · financialcontent.com/article/bizwire-2026-10-5-s | verified |
| Shift Bioscience virtual cell and cell-ageing clock platform (SB000 discovery)Chips, models and products | Shift Bioscience's AI "virtual cell" and ageing-clock tools, used to find single genes that may make old cells behave younger. live (in-house discovery platform)Key facts 9 Jun 2025: platform identified SB000, a single-gene target that reverses cellular ageing as measured by DNA methylation clocks in multiple cell types; stated "twice the magnitude of cell methylome rejuvenation compared to OSKM"; preprint bioRxiv 10.1101/2025.06.05.657370; platform model not named; access: in-houseMade by Shift Bioscience · 2025-06-09 (first named target announced) | Cambridge (via Shift Bioscience) | shiftbioscience.com/news/shift-bioscience-identi | one source |
| Tolemy BioCompanies | Cambridge start-up building "Orbit", a virtual-cell platform with AI research agents that help scientists read and steer cell behaviour for therapies.Key facts €1.4m pre-seed (2026-05-12)Profile
Labs Round 8 deep profile (31 Oct 2026), as the researchers wrote it; confidence: one source. Sources | Cambridge (also Barcelona) | tolemy.bio/ · tech.eu/2026/05/12/tolemy-bio-lands-eur14m-to-po · tolemy.bio/team · tolemy.bio/platform | one source |
| Trillion Gene AtlasChips, models and products | Basecamp Research's plan to sequence a trillion genes from nature to train its AI models. announced (data programme)Key facts Announced 18 Mar 2026 (trade headlines listed on the lab's own news page; the lab's PDF release could not be opened). Treat scale claims as unverified until the release is read.Made by Basecamp Research · 2026-03-18 | Clerkenwell Workshops, 27 Clerkenwell Close, London (EC1R 0AT) (via Basecamp Research) | basecamp-research.com/news-research | one source |