Query and lookup resources for cellular-agriculture AI — sequence and expression repositories,
livestock atlases, protein and structure databases, pathway and metabolic-model collections,
spectral libraries, and the directories that map the field. Each entry links to its canonical
home; full descriptions live in the canonical Databases.md.
GEO is the NCBI's public functional genomics data repository, accepting array- and sequence-based data submissions from researchers worldwide and currently hosting millions of samples across hundreds of thousands of studies. Coverage spans bulk and single-cell transcriptomics, methylation, ChIP-seq, ATAC-seq, and other genome-wide assays in essentially every model organism plus many livestock species (bovine, porcine, ovine, gallus, salmonids). For cellular agriculture, GEO is a primary source of training data for ML models of cell-line-specific expression responses, comparative transcriptomics across muscle / adipose / fibroblast lineages, and condition-dependent media responses. Programmatic access via GEOquery (R), GEOparse (Python), and pysradb; raw data flows through the linked SRA repository.
The Sequence Read Archive is NCBI's public repository for high-throughput raw sequencing data, hosting reads from essentially every major sequencing platform (Illumina, PacBio, Oxford Nanopore, BGI, Element, Singular Genomics) across petabases of stored data. Coverage spans bulk and single-cell RNA-seq, ATAC-seq, ChIP-seq, whole-genome and exome sequencing, and metagenomics across thousands of species, including the major cell-ag livestock (bovine, porcine, ovine, gallus, salmonids). For cellular agriculture, SRA is the primary source of raw read data for re-analyzing public studies with custom pipelines — re-aligning bovine satellite cell RNA-seq to a more recent genome build, recomputing single-cell embeddings under different normalization, or extracting novel features for ML models that the original analysis did not surface. Programmatic access via sra-tools (prefetch / fasterq-dump) and pysradb; SRA data is mirrored on AWS Open Data and Google Cloud for cloud-based processing.
GenBank is the NIH's annotated collection of all publicly available nucleotide sequences, comprising over a trillion bases across millions of distinct organisms. It serves as the primary deposition site for newly sequenced genes, transcripts, genomes, and constructs, with entries flowing into the broader NCBI ecosystem (RefSeq, BLAST, Genome, Nucleotide, Protein). For cellular agriculture, GenBank is the authoritative source for canonical gene sequences of cell-ag-relevant species — bovine MYOD / IGF1 / FGF2; chicken Pax3/Pax7; salmon myogenic regulatory factors; and the cytokines, growth factors, and signaling molecules used in serum-free media formulations. Programmatic access via Entrez (EFetch / ESearch), the NCBI Datasets CLI, and BLAST for sequence similarity search.
Ensembl is the principal vertebrate genome browser and annotation database, jointly maintained by EMBL-EBI and the Wellcome Sanger Institute, providing automated genome annotation pipelines, comparative genomics, and variation data for >300 vertebrate species. Critically for cellular agriculture, Ensembl hosts high-quality genome assemblies for all major cell-ag livestock — Bos taurus (cow), Sus scrofa (pig), Gallus gallus (chicken), Ovis aries (sheep), Salmo salar (Atlantic salmon), Oncorhynchus mykiss (rainbow trout), and others — with gene models, regulatory features, orthology mappings to human, and tissue-specific expression annotations. The Ensembl REST API, Compara perl toolkit, and BioMart enable programmatic large-scale data extraction for ML pipelines. Ensembl Genomes (a sister site) provides parallel data for bacteria, protists, fungi, plants, and invertebrates relevant to precision-fermentation cell-ag.
ArrayExpress, now part of the broader BioStudies platform at EMBL-EBI, is a public archive of functional genomics experiments — array-based and sequence-based, bulk and single-cell — covering transcriptomics, epigenomics, and related assays. Many studies are deposited simultaneously in GEO and ArrayExpress, but a meaningful subset (particularly from EU-based labs) is unique to ArrayExpress, making it a complementary source to GEO for comprehensive data mining. For cellular agriculture, ArrayExpress is a useful secondary index when searching for European-led work on livestock cell biology, comparative myogenesis, and cell-line characterization. Programmatic access via the BioStudies REST API.
The Farm Animal Genotype–Tissue Expression project — an international consortium developing GTEx-style multi-tissue resources for livestock species. Hosts the umbrella project portal plus links to per-species sub-portals (PigGTEx, ChickenGTEx, CattleGTEx) and shared infrastructure (TWAS-Server, PigBiobank). The canonical entry point for livestock systems-genetics work. Companion to Papers.md ref #134 (Fang et al. 2025, Nature Genetics).
Chicken sub-portal of FarmGTEx, providing multi-tissue genetic-regulation maps across chicken tissues — the first GTEx-style resource for a non-mammalian amniote livestock species, directly useful for cultivated-chicken cell-line engineering work. Companion to Papers.md ref #136 (Guan et al. 2025, Nature Genetics). Processed-data deposit: Zenodo 10.5281/zenodo.14902956. Cross-referenced from Datasets/Chicken.md.
Cattle sub-portal of FarmGTEx, providing bulk and single-cell multi-tissue expression atlases for cattle (Bos taurus) — directly relevant to cultivated-beef cell-line characterization and engineering. Companion to Papers.md ref #137 (Han et al. 2025, Nature Genetics, the single-cell atlas paper) and Papers.md ref #192 (Liu et al. 2022, Nature Genetics, the foundational CGTEx multi-tissue regulatory-variants atlas). Han 2025 deposits: NCBI BioProject PRJNA1119173; Zenodo 10.5281/zenodo.16572998 and 10.5281/zenodo.15721498. Liu 2022 deposits: Zenodo 10.5281/zenodo.6510550; pipeline at shuliliu/cattleGTEx. Cross-referenced from Datasets/Cow.md.
Sheep sub-portal of FarmGTEx, providing multi-tissue regulatory-effects maps across sheep (Ovis aries) tissues — extending the FarmGTEx pattern from cattle / pig / chicken into the small-ruminant lineage. Companion to Papers.md ref #138 (Gong et al. 2025, bioRxiv). Pipeline: FarmGTEx/SheepGTEx-Pipeline-v0. Raw deposits: NCBI BioProjects PRJNA1198671, PRJNA1304012. Cross-referenced from Datasets/Sheep.md.
Server providing transcriptome-wide association study (TWAS) analyses across the FarmGTEx tissue / sample matrix — links expression-genetics signals to complex-trait associations across livestock species. Companion infrastructure to the species-specific FarmGTEx sub-portals above.
Trait-data biobank coordinated with PigGTEx, integrating phenotypic, genotypic, and expression data on diverse pig traits — complementary to PigGTEx-Portal's molecular-QTL focus. Companion to Papers.md ref #139 (Zeng et al. 2024, Nucleic Acids Research). Cross-referenced from Datasets/Pig.md.
A global research consortium and resource network for functional annotation of farmed-animal genomes, paralleling ENCODE for human/mouse but focused on livestock species. Predates FarmGTEx and provides the broader functional-annotation substrate (chromatin state, regulatory elements, etc.) that FarmGTEx expression-genetics work builds on. Companion to Papers.md ref #144 (Clark et al. 2020, Genome Biology).
The FAANG consortium's aquaculture programme — Advancing European Aquaculture by Genome Functional Annotation — generating genome-wide functional annotation maps for six aquaculture species important to European aquaculture (Atlantic salmon, rainbow trout, European sea bass, gilthead sea bream, common carp, turbot). The data hub hosts the project's open releases (55 datasets at time of curation, spanning RNA-seq, ATAC-seq, and ChIP-seq). A FAANG functional-annotation substrate for aquaculture-cell-line work, paralleling FarmGTEx for terrestrial livestock. Cross-referenced from Datasets/Fish.md.
The FAANG consortium's cattle functional-annotation project, generating maps of functionally active genomic features in cattle (Bos taurus). The data hub hosts the open releases (4 datasets at time of curation). Complementary to CattleGTEx for cell-ag-relevant cattle-cell engineering and atlas work — particularly useful as a regulatory-element substrate for cultivated-beef cell-line characterisation. Cross-referenced from Datasets/Cow.md.
The FAANG consortium's pig + chicken project — the regulatory GENomE of SWine and CHicken: functional annotation during development. The data hub hosts the open releases (19 datasets at time of curation, covering pig and chicken functional genomes during development). Substrate for cultivated-pork and cultivated-chicken developmental-biology and lineage-engineering work; complementary to PigGTEx-Portal and ChickenGTEx-Portal. Cross-referenced from Datasets/Pig.md and Datasets/Chicken.md.
UniProt is the central knowledge base for protein sequence and functional information, jointly maintained by the SIB, EMBL-EBI, and PIR. It comprises SwissProt (manually curated, ~570K entries) and TrEMBL (computationally derived, ~250M entries), with each entry providing sequence, functional annotations, GO terms, post-translational modifications, subcellular localization, cross-references to structure databases, and supporting literature. For cellular agriculture, UniProt is the canonical reference for growth factors, signaling proteins, transcription factors, and metabolic enzymes — including the recombinant FGF2, IGF1, TGF-β, and HGF variants and their orthologs across bovine, porcine, chicken, and fish species used in cultivated-meat media. Programmatic access via the UniProt REST API, SPARQL endpoint, and downloadable XML / FASTA / TSV bulk files.
The AlphaFold Database is a freely accessible repository of predicted 3D protein structures generated by DeepMind's AlphaFold model, jointly hosted by DeepMind and EMBL-EBI. It contains predicted structures for >214 million proteins covering essentially the entirety of UniProt's reference proteomes, including all proteins in human, mouse, bovine, porcine, chicken, salmon, and many other species relevant to cellular agriculture. Each entry provides predicted Cartesian coordinates, per-residue confidence scores (pLDDT), and predicted aligned error (PAE) matrices for assessing inter-domain confidence. For cellular agriculture, AlphaFold DB is the substrate for structure-based ML on engineered growth factors, recombinant signaling molecules, and stable scaffold proteins for media formulation. Programmatic access via the AlphaFold API and bulk PDB / CIF downloads through Google Cloud.
The Protein Data Bank is the worldwide repository of experimentally determined biomolecular structures, jointly maintained by RCSB (US), PDBe (UK), and PDBj (Japan), containing >220,000 entries from X-ray crystallography, NMR, cryo-EM, and other experimental techniques. Each entry provides atomic coordinates, experimental data, validation metrics, and metadata including bound ligands, mutations, and binding partners. For cellular agriculture, PDB serves as the canonical complement to AlphaFold's predicted structures — providing experimentally validated structures of growth factors, cytokines, receptors, and enzymes that can be used as ground truth for protein engineering ML models or as input templates for structural modeling pipelines. Programmatic access via the RCSB REST API, GraphQL endpoint, and mmCIF / PDB file downloads; structure similarity search via the integrated FoldSeek and DALI services.
InterPro is an integrative classification of protein families, domains, and functional sites, hosted by EMBL-EBI and pulling annotations from 13 member databases (Pfam, SMART, PROSITE, CATH-Gene3D, PANTHER, and others). It maps every UniProt entry to InterPro entries, providing inferred functions, domain architectures, and evolutionary classifications. For cellular agriculture, InterPro is useful for identifying functional domains in recombinant proteins targeted for media supplementation, locating conserved motifs across species orthologs (e.g. ensuring an engineered bovine FGF2 variant retains the heparin-binding site of human FGF2), and predicting the functional impact of mutations in protein engineering campaigns. Programmatic access via the InterPro REST API, downloadable bulk files, and the InterProScan command-line tool for annotating new sequences against all member databases at once.
The Human Protein Atlas is a large-scale Swedish-led effort to map the expression and localization of every human protein-coding gene at protein and RNA level across tissues, organs, single cells, subcellular compartments, and pathological contexts. It provides immunohistochemistry images, single-cell transcriptomics summaries, secretome and membrane-proteome annotations, and computed expression scores across 50+ normal tissues, ~70 cancer types, and dozens of cell lines. For cellular agriculture, HPA is useful for identifying which growth factors and receptors are natively expressed in target tissues (skeletal muscle, adipose, dermal fibroblast) so that media formulations and engineered cells reflect physiologically relevant signaling environments; despite the human focus, many findings translate to livestock species through orthology. Programmatic access via per-protein TSV downloads and a REST API.
STRING is a database of known and predicted protein-protein interactions, integrating experimental, computational, text-mining, and co-expression evidence into weighted interaction networks for >14,000 species, hosted by the von Mering lab (University of Zurich / SIB). Each interaction has an associated confidence score and supporting evidence type, allowing downstream analyses to filter for high-confidence physical interactions or to include broader functional associations. For cellular agriculture, STRING is useful for graph-based ML on cell-signaling networks, identifying co-regulated modules in myogenesis or adipogenesis pathways, and constructing prior-knowledge inputs for ML models that benefit from biological structure (graph neural networks, network-regularized regression). Programmatic access via the STRING REST API, the stringdb R package, and a Cytoscape app.
PRIDE (PRoteomics IDEntifications database) is the world's largest public repository of mass-spectrometry-based proteomics data, hosted at EMBL-EBI and integrated into the international ProteomeXchange consortium. It hosts raw spectra, peptide and protein identifications, post-translational modification data, and quantitative analyses from tens of thousands of studies, covering bulk and single-cell proteomics, phosphoproteomics, glycoproteomics, and clinical mass-spec workflows. For cellular agriculture, PRIDE provides training data for ML models predicting cell-line-specific protein expression, identifying biomarkers for differentiation states, and validating predicted protein expression under varying media conditions. Programmatic access via the PRIDE REST API and ProteomeXchange consortium endpoints.
Cellosaurus is a knowledge resource on cell lines maintained by the Swiss Institute of Bioinformatics (SIB) as part of the ExPASy infrastructure, cataloging over 170,000 cell lines across human, rodent, bovine, porcine, chicken, fish, and other species. Each entry provides a standardized Cellosaurus identifier (CVCL_XXXX), parent / derivative relationships, species of origin, tissue / disease context, donor demographics, known karyotype and STR profiles, authenticity status (notably flagging misidentified or contaminated lines), and cross-references to physical-repository deposits (ATCC, ECACC, RIKEN, JCRB, and others). For cellular agriculture, Cellosaurus is the canonical resource for resolving cell-line identity in published cultivated-meat literature and for sourcing background on bovine satellite cells, immortalized fibroblast lines, fish embryonic stem cell lines, and other species-specific cell lines under active investigation. Data is freely downloadable as flat OBO / TXT files and queryable via a REST API.
The Human Cell Atlas is an international consortium effort to construct comprehensive single-cell reference maps of every human cell type, integrating contributions from hundreds of labs across thousands of donors and dozens of tissues. The HCA Data Portal hosts harmonized scRNA-seq, snRNA-seq, scATAC-seq, and spatial transcriptomics datasets with consistent metadata schemas, cell-type ontologies, and processing pipelines, totaling tens of millions of annotated cells. For cellular agriculture, HCA is a primary reference for understanding cellular heterogeneity within target tissues (distinguishing satellite cells from interstitial fibroblasts in skeletal muscle, mapping adipocyte subtypes in white adipose tissue), validating differentiation trajectories in cultivated cells, and benchmarking cell-state engineering interventions against in vivo references. Programmatic access via the HCA Data Portal API and the integrated cellxgene Census.
CZ CELLxGENE is the Chan Zuckerberg Initiative's platform for exploring and analyzing public single-cell data, providing harmonized scRNA-seq and multi-omics datasets aggregated from hundreds of studies (including Human Cell Atlas data) under consistent metadata standards. The CELLxGENE Census provides programmatic access to tens of millions of harmonized cells across human and mouse, exposed through a TileDB-backed Python / R API that supports fast queries by gene, tissue, disease, or assay. For cellular agriculture, cellxgene is the most practical source for assembling cross-study training sets (e.g. all skeletal muscle stem cells in the Census) for ML models of cell-type classification, perturbation prediction, and trajectory inference. Programmatic access via the cellxgene-census Python and R packages.
KEGG (Kyoto Encyclopedia of Genes and Genomes) is one of the most widely used pathway databases, maintained by the Kanehisa Laboratory at Kyoto University, providing manually curated maps of >500 reference metabolic, signaling, and disease pathways. KEGG PATHWAY entries link genes, proteins, compounds, and reactions across thousands of organisms, with each pathway available as a structured KGML XML file suitable for ML pipelines that need graph structure. For cellular agriculture, KEGG is the canonical reference for central metabolism (glycolysis, TCA cycle, oxidative phosphorylation, fatty-acid biosynthesis) that drives proliferation in cultivated cells, signaling pathways (Wnt, FGF, IGF, TGF-β) that govern differentiation, and amino-acid biosynthesis pathways relevant to serum-free media design. Programmatic access via the KEGG REST API (free for non-commercial use; commercial licensing applies).
Reactome is a free, open-source, curated database of human (and increasingly other species) biological pathways, jointly developed by the Ontario Institute for Cancer Research, EMBL-EBI, and other partners. It models pathways as hierarchically organized reactions with explicit participants, modifiers, and regulatory relationships, supporting pathway enrichment analysis and integration with expression data. For cellular agriculture, Reactome's strengths are its detailed signaling-pathway annotations (myogenic regulatory factor cascades, FGF / IGF / Wnt signaling) and its open data model — entries are freely downloadable as BioPAX, SBML, SBGN, and tab-delimited formats, enabling unrestricted use in ML pipelines without the commercial licensing constraints of KEGG. Programmatic access via the Reactome ContentService REST API and a Cypher-queryable Neo4j graph database.
WikiPathways is a community-curated database of biological pathways under an open data model (CC0 license), hosted by Maastricht University and the Gladstone Institutes, providing >3,000 curated pathways across human, mouse, and a range of other species. Pathways are stored in GPML (a custom XML format) but exported in BioPAX, SBML, and other standards; the open editorial model means coverage is broader and more uneven than KEGG / Reactome but includes useful cell-ag-adjacent pathways (myogenesis, adipogenesis, livestock-specific metabolism). Programmatic access via a Web Service API, the rWikiPathways R package, and a Cytoscape app for direct pathway import.
HMDB is a freely available, comprehensive database of human metabolites and their associated biological roles, maintained by the Wishart Lab at the University of Alberta. It contains >220,000 entries spanning endogenous metabolites, drugs, food components, and environmental compounds, with each entry providing chemical / physical properties, biofluid concentrations, biological functions, metabolic-pathway memberships, and analytical detection methods. For cellular agriculture, HMDB is useful for understanding native human metabolite ranges that cultivated cells should produce or consume, identifying candidate small molecules for media supplementation, and as a substrate for ML predicting metabolic responses to media changes. Programmatic access via downloadable XML and SDF files and a REST API for selected queries.
BMDB is a comprehensive resource of metabolites found in beef cattle (Bos taurus), maintained by the Wishart Lab at the University of Alberta — the same group that hosts HMDB, FooDB, and DrugBank, providing consistent schema and analytical methodology across all four resources. It catalogs tens of thousands of entries describing endogenous metabolites, dietary nutrients, and exogenous compounds detected in bovine biofluids (serum, milk, urine) and tissues (muscle, liver, adipose), with each entry providing chemical / physical properties, biofluid concentrations, biological roles, pathway memberships, and analytical methods of detection. For cultivated beef development, BMDB is the closest available reference for serum metabolite profiles, muscle and adipose tissue metabolomes, and feedstock-derived compound landscapes — directly relevant to serum-free media formulation, growth medium optimization, and flavor-precursor identification in cultivated meat products.
BRENDA (BRaunschweig ENzyme DAtabase) is the most comprehensive collection of enzyme information available, maintained at the Technische Universität Braunschweig, covering >90,000 enzyme entries with detailed kinetic parameters, substrate specificities, inhibitors, optimal reaction conditions, and reaction mechanisms aggregated from primary literature. Each entry provides Km, Vmax, Kcat, Ki, optimal pH and temperature, organism-specific variations, and links to associated metabolic pathways. For cellular agriculture, BRENDA is the canonical source of enzyme kinetic data needed for kinetic models of cell metabolism, flux-balance analysis of metabolic engineering interventions, and ML models predicting media component effects on metabolic flux. Programmatic access via the BRENDA SOAP API and downloadable text files for academic use.
SMPDB is the Wishart lab's curated database of small-molecule pathways, with over 600 interactive visual pathway maps spanning human metabolism, signaling, drug action, disease processes, and physiological functions. Each pathway entry provides labeled compounds, enzymes, transporters, and reactions in a custom interactive viewer, with links to HMDB / DrugBank / UniProt cross-references. For cellular agriculture, SMPDB serves as a visualization-rich complement to KEGG and Reactome for understanding the metabolic context of media supplements, growth factors, and signaling pathways relevant to cultivated cells. Free open access; data exportable as SBML, BioPAX, KGML, and PWML.
BioCyc is an SRI International–maintained collection of >20,000 organism-specific Pathway / Genome Databases (PGDBs), built on the MetaCyc reference pathway database. MetaCyc itself catalogs >3,000 experimentally elucidated metabolic pathways across all domains of life, with >18,000 reactions and >19,000 metabolites; BioCyc PGDBs extend this to specific organisms with computationally predicted pathways. For cellular agriculture, BioCyc is the canonical reference for non-mammalian metabolism — particularly the yeast and bacterial PGDBs used in precision-fermentation alt-protein work, and emerging livestock-species PGDBs that complement species-specific GEMs. MetaCyc is freely accessible at https://metacyc.org/; BioCyc has tiered access with substantial free use plus subscription-based bulk download.
A repository of >75 manually curated, BiGG-standardized genome-scale metabolic models covering bacteria, archaea, fungi, and several eukaryotes, hosted at UCSD (Palsson lab). Each model is provided in SBML, MATLAB .mat, and JSON formats, with standardized reaction / metabolite IDs (BiGG nomenclature) that cross-link to MetaNetX, KEGG, and ChEBI. The canonical starting point for any constraint-based modeling work (King et al. 2016, Nucleic Acids Research).
A free, open-source repository of mathematical models of biological and biomedical systems, hosted by EMBL-EBI, containing thousands of curated SBML models spanning metabolism, signaling, cell-cycle, immunology, and pharmacokinetics. Models are versioned, peer-reviewed via the Curation Service, and accessible programmatically via a REST API. Endorsed companion to the Talk2Biomodels agent (Papers.md ref #50).
MassBank is the world's first open-access mass-spectrometry spectral database, hosted as a federation of European, Japanese, and US instances with >400,000 spectra contributed by >50 institutions. Each entry provides the recorded mass spectrum, instrumental parameters, fragmentation conditions, and the precursor compound identity (InChI, SMILES, structure). For cellular agriculture, MassBank is foundational reference data for identifying volatile and non-volatile metabolites in cultivated tissues, characterizing off-flavors, and analyzing spent media — particularly when paired with MS-DIAL, MZmine 3, or SIRIUS for automated annotation. Programmatic access via REST API; bulk records available via Git.
The US National Institute of Standards and Technology's reference compendium of chemistry data, covering >100,000 compounds with thermochemical data (heats of formation, enthalpies), gas chromatography retention indices, mass spectra, infrared / UV / NMR spectra, and ion energetics. For cellular agriculture, the NIST WebBook is the canonical reference for confirming media component identity via spectroscopic comparison (matching unknown GC-MS peaks against the NIST EI library is standard practice) and for thermodynamic calculations in metabolic engineering. Free open-web access; the bundled NIST 23 EI library (commercial) is the industry-standard companion for GC-MS workflows.
MetaboLights is a public open-access repository for metabolomics experiments hosted at EMBL-EBI — the metabolomics analog of GEO (transcriptomics) or PRIDE (proteomics). Each study deposit includes raw MS / NMR data, processed datasets, sample / treatment metadata in ISA-Tab format, and standardized study-protocol annotations. For cellular agriculture, MetaboLights hosts experiments directly relevant to flavor metabolomics, fermentation chemistry, and cell-culture spent-media analysis — a primary source of training data for sensory-prediction ML and benchmark datasets for analytical pipeline validation. Programmatic access via REST API; bulk download via FTP.
The Global Natural Products Social Molecular Networking platform from the Dorrestein lab at UCSD provides community-curated MS/MS reference spectral libraries — the database aspect listed here — alongside Feature-Based Molecular Networking, Ion Identity Molecular Networking, and analog-search tooling, which together make it primarily an analysis platform. Full entry and platform-level capabilities are in Software.md / Mass Spectrometry & Chemometrics; this dual-listing surfaces it as a spectral database for readers browsing Databases.md.
ChEMBL is a manually curated database of bioactive small molecules with drug-like properties, maintained at EMBL-EBI, integrating compound structures, biological activity data, target information, and mechanism-of-action annotations from medicinal chemistry literature, patents, and screening assays. It contains >2.5M unique compounds, ~16M activity measurements, and >15K molecular targets. For cellular agriculture, ChEMBL is useful for identifying small-molecule alternatives to recombinant growth factors (e.g. small-molecule Wnt agonists, MEK inhibitors used in differentiation protocols), screening compound libraries for cell-ag-relevant bioactivities, and as training data for ML models predicting small-molecule effects on cellular pathways. Programmatic access via the ChEMBL REST API, downloadable SQL dumps for full local instances, and the chembl-webresource-client Python client.
PubChem is NIH's open chemistry database, the largest public repository of chemical structures and bioactivity data, hosting >120M compounds, >330M substances, and >1.4M bioassays. Each compound entry provides structure, computed physico-chemical properties, synonyms, classifications, biological activities, patents, literature references, and safety / toxicology information. For cellular agriculture, PubChem is the broadest practical reference for media components, supplements, and small-molecule modulators — including industrial-grade ingredients without the literature-grade coverage of ChEMBL. Programmatic access via the PubChem PUG REST API, PUG-View for richer record summaries, and SQL queries over the PubChem data warehouse.
DrugBank is a comprehensive bioinformatics / cheminformatics resource maintained by the Wishart lab at the University of Alberta, cataloging >13,000 drug entries — including ~2,200 FDA-approved small molecules, ~340 FDA-approved biologics, ~93 nutraceuticals, and >5,000 experimental drugs — with detailed chemistry, pharmacology, pharmaceutical, and pharmacological-target data. For cellular agriculture, DrugBank is useful for identifying small-molecule alternatives to recombinant growth factors (many drug-like compounds modulate cell-signaling pathways relevant to differentiation and proliferation), for cross-referencing media supplements against known pharmacological activities, and as a sister resource to HMDB / BMDB / SMPDB in the Wishart-lab ecosystem. Programmatic access via REST API; downloads require academic registration.
ChemSpider is the Royal Society of Chemistry's free chemical structure search platform, aggregating ~88 million unique chemical structures from over 270 data sources including PubChem, ChEBI, FDA, EPA, vendor catalogs, and primary literature. Each entry provides identifiers (InChI, SMILES, CAS, IUPAC name), structure depiction, computed physicochemical properties, and source provenance. For cellular agriculture, ChemSpider serves as a comprehensive fallback when ChEMBL or PubChem don't cover a particular media component or flavor precursor — particularly useful for obscure natural products or vendor-specific compounds.
ChEBI (Chemical Entities of Biological Interest) is EMBL-EBI's curated ontology of biologically relevant chemical entities, with ~200,000 annotated compounds spanning metabolites, drugs, nutrients, and toxins. Each entry includes structures, ontological classifications (biological role, chemical class), cross-references to other databases, and contributor-curated literature. For cellular agriculture, ChEBI is the structured chemical ontology underlying many pathway databases (KEGG, Reactome, BioCyc), making it useful as a controlled vocabulary for annotating compound roles in cell-ag-relevant metabolic networks and for cross-database identifier mapping. Programmatic access via REST API and SOAP web services.
T3DB is the Wishart lab's curated database of toxic-exposome compounds, cataloging ~3,500 environmental contaminants, pollutants, industrial chemicals, and endogenous toxic metabolites along with their molecular targets and biological effects. Each entry provides chemical / physical properties, biological mechanism of action, toxicity references, and analytical methods of detection. For cellular agriculture, T3DB is useful for identifying potential off-flavor or off-odor compounds in cultivated tissues (some volatile toxins overlap with food chemistry — aldehydes, aromatic compounds, sulfides), for media-safety screening, and for understanding adulterant detection in regulatory contexts.
FlavorDB is a comprehensive database of flavor molecules developed by the Complex Systems Lab at IIIT-Delhi (Bagler group). The original release (Garg et al. 2018, Nucleic Acids Research) catalogued 25,595 flavor molecules across 936 ingredients, integrating data from FooDB, BitterDB, SuperSweet, and Flavornet. FlavorDB2 (Goel et al. 2024, Journal of Food Science) extends coverage with regulatory status, consumption statistics, taste/aroma threshold values, reported food uses, and synthesis pathways. For cellular agriculture, FlavorDB2 is a key reference for connecting volatile compounds detected in cultivated tissues (e.g., GC-MS of cultivated pork fat or salmonid muscle) to known flavor characteristics, organoleptic thresholds, and food applications.
BitterDB is a curated database of bitter compounds and bitter taste receptors (T2Rs), maintained at the Hebrew University of Jerusalem (Niv group). Contains hundreds of compounds with chemical properties, sources, and human T2R-binding data; companion tools (BitterPredict, BitterMatch, BitterMasS) leverage the data for predictive modeling. For cellular agriculture and alt-protein work, BitterDB is essential for identifying and predicting bitter off-flavors — a primary sensory barrier in plant-based protein products and a potential concern in cultivated tissues with elevated metabolite content.
Canonical references: Papers.md #99 (Wiener et al. 2012, NAR) for the original release; Papers.md #100 (Dagan-Wiener et al. 2019, NAR) for the 2019 update; Papers.md #101 (Ziaikin et al. 2025, NAR) for the 2024 update. Related ML / analytical papers from the same group are in matrix refs #102, #103, #104, #105 and Sensory & Flavor Reference Work ref #106.
A comprehensive database of insect and animal pheromones, semiochemicals, and other behaviorally active compounds, maintained by Ashraf El-Sayed. Contains tens of thousands of compounds with biological activity, taxonomic distribution, and analytical data (GC retention indices, mass spectra). For cellular agriculture, Pherobase has indirect but useful relevance — many flavor-active compounds in meat / dairy / fish are downstream of pheromonal or species-specific biochemistry, and the database serves as a reference for distinguishing odor compounds that evolved for animal-animal signaling versus those that drive food-grade flavor perception.
Flavornet, maintained at Cornell by Acree and Arn, is a database of GC olfactometry (GC-O) results — compound identities keyed to Kováts retention indices on standard GC columns plus consensus odor descriptors from sensory panels. The original open-access reference for matching sniff-port observations to molecular identities in GC-O work. For cellular agriculture, Flavornet is the canonical lookup table for connecting volatile compounds detected in cultivated tissues, fermented alt-protein products, and spent media to their human-perceived odor profiles.
A long-established industry reference platform for the flavor, fragrance, food, and cosmetics industries, providing odor descriptors, organoleptic properties, regulatory status, and supplier information for thousands of aroma chemicals and natural extracts. While maintained by the F&F industry rather than as an academic resource, GoodScents is widely cross-referenced in flavor research papers and is the de-facto industry-standard descriptor source for many compounds not formally characterized in academic odor-threshold compendia.
FooDB is a large open database of food constituents, chemistry, and biology — covering both macronutrients and the thousands of micro-constituents (flavor and aroma compounds among them) found in foods, with chemical, taxonomic, and concentration data. It is one of the upstream sources integrated into FlavorDB. For cellular agriculture, FooDB is a reference for the compound composition of conventional foods — a comparison baseline when characterizing the metabolite and flavor profile of cultivated tissues and alt-protein products.
PISCES organizes seafood data by phylogenetic relationships, grouping species taxonomically — Salmo salar (Atlantic salmon) and Salmo trutta (sea trout) sit together under genus Salmo. Each species entry compiles cell-line availability, nutritional data, and volatile compounds relevant to cultivated- and plant-based-seafood development. Distributed as an AirTable base linked from the GFI landing page. The canonical taxonomic reference for cultivated-seafood candidate species. Companion to ATLAS below, which groups the same species by culinary archetype rather than phylogeny.
ATLAS groups seafood species into culinary archetypes — Salmo salar under "salmon" and Salmo trutta under "trout" — and ranks archetypes across sustainability, animal-welfare, public-health, and US-market-size metrics to help prioritize candidate species for alternative-seafood development. Accessed via the GFI landing page, with a ranking tool hosted on mybinder.org. The species-prioritization complement to PISCES's taxonomic indexing.
AGRIS is the United Nations Food and Agriculture Organization's global agricultural research index, providing access to over 12 million bibliographic records covering food, agriculture, environment, and rural development since 1970. Each record includes title, authors, abstract, subject keywords (drawn from the multilingual AGROVOC ontology), and full-text links where available. For cellular agriculture, AGRIS is the most comprehensive free index for international agricultural research — useful for feed formulation, plant-based media components, livestock nutrition, and crop-genomics work that intersects with cell-ag.
The National Agricultural Library Search is the USDA's federated search platform across its digital collections, including PubAg (agricultural research literature), the NAL Catalog, and other USDA-curated resources. For cellular agriculture, NAL Search is the canonical entry point for USDA-funded research on animal nutrition, feed efficiency, livestock genetics, food technology, and food safety — relevant context for cultivated-meat scaling economics, regulatory engagement, and feed-stream cost analysis. Free open access; AGRICOLA-derived records are included alongside open-access NAL collections.
GFI's curated collection of alternative-protein research resources spanning plant-based, cultivated, and fermentation-derived proteins, maintained by its science and technology team. GFI's nearest analogue to CAAIL's own curation, though not AI-scoped or version-controlled.
A community-maintained index of biological databases, grouped by type — meta-databases, model-organism, nucleic-acid, protein, pathway, taxonomic, and more. A broad navigation layer for biological data resources beyond those catalogued here.
A community-maintained list of websites and databases of chemical information. Complements the Chemistry & Compound Databases section above as a broader navigation layer for chemistry data resources.
The University of Arkansas Libraries' research guide for food science, listing relevant databases, journals, and statistical resources — a curated entry point into food-science literature and data infrastructure.
GFI's global directory of researchers working on plant-based meat, cultivated meat, and fermentation. Complementary to CAAIL: GFI catalogues people and labs, CAAIL catalogues their outputs.
GFI APAC's regional B2B directory of ingredient and equipment suppliers, pilot plants, consultants, and other supply-chain partners across the Asia-Pacific cell-ag ecosystem.
GFI's live tracker of countries where cultivated meat has regulatory approval for sale, with associated companies and product formats — the kind of dynamic regulatory data CAAIL links to rather than duplicates.
A web-based tool from the Global Alliance for Improved Nutrition (GAIN), Johns Hopkins University, Cornell University, and the FAO that compiles and visualizes national- and subnational-level food-systems data to inform food policy (introduced by Fanzo et al. 2020, Nature Food, 10.1038/s43016-020-0077-y). Organized around three pillars — Describe, Diagnose, Decide — it brings together over 300 indicators from 50+ public and private sources (UN agencies, the World Bank, CGIAR, Euromonitor) spanning agricultural production, food availability and affordability, diets and nutrition, livelihoods, climate, environment, resilience, and governance, plus the external drivers shaping them; the Diagnose view scores 39 indicators per country and the Decide view offers 87 evidence-based policies and actions. For cellular agriculture, it supplies macro food-systems context (dietary, sustainability, and policy framing) that situates cultivated-meat and alternative-protein interventions, complementary to CAAIL's molecular, dataset, and tooling resources.
The Prediction Center hosts CASP (Critical Assessment of Structure Prediction) — the long-running community experiment and results tracker for protein-structure-prediction methods, run biennially since 1994, providing an independent assessment of prediction methods against experimental structures. AlphaFold2's breakthrough accuracy was independently assessed here at CASP14 (2020). For cellular agriculture, CASP is the canonical reference point for judging structure-prediction tools applied to growth factors, recombinant ECM and scaffold proteins, and other engineered proteins in cultivated-meat workflows.
Genentech's open leaderboard hosted on Hugging Face Spaces, tracking model performance on the 100-task CompBioBench v1 computational-biology benchmark. Provides a vendor-neutral, continuously-updated comparison point for cell-ag teams evaluating LLMs and agents on bioinformatics tasks before deploying them downstream. The underlying benchmark data — questions plus the BAM/FASTQ/H5AD/MTX/TSV bioinformatics artifacts each task operates over — is catalogued separately at Datasets/Benchmarks.md / CompBioBench v1 (canonical home: the Genentech/compbiobench-data-v1 Hugging Face Dataset). Companion to Papers.md ref #150 (Nair et al. 2026), the bioRxiv methods paper describing the benchmark.
The official results tracker at lastexam.ai for Humanity's Last Exam, ranking frontier models on the 2,500-question closed-ended benchmark at the frontier of human knowledge. The companion benchmark data is catalogued at Datasets/Benchmarks.md / Humanity's Last Exam; paired with Papers.md ref #158 (Phan et al. 2026). A continuously-updated ceiling check on frontier reasoning for cell-ag teams choosing a general-purpose model.
The live leaderboard on Hugging Face Spaces for the reasoning-focused MMLU-Pro benchmark, tracking model accuracy across its 14 disciplines. The companion benchmark data is catalogued at Datasets/Benchmarks.md / MMLU-Pro; paired with Papers.md ref #157 (Wang et al. 2024). A vendor-neutral general-reasoning comparison point for sizing up models before cell-ag-specific evaluation.
The live leaderboard hosted at proteingym.org for the ProteinGym variant-effect benchmark — separate substitution and indel boards covering supervised and zero-shot model categories. The companion benchmark data is catalogued at Datasets/Benchmarks.md / ProteinGym; paired with Papers.md ref #148 (Notin et al. 2023). The dominant variant-effect leaderboard in protein-engineering ML — directly relevant to any cell-ag protein-engineering work (growth factors, scaffolds, recombinant ECM proteins) selecting a protein language model.
The official leaderboard at swebench.com tracking agent and model performance across the SWE-bench leaderboards — the Full set plus the Verified, Lite, Multilingual, and Multimodal variants — by the percentage of real GitHub issues resolved with a test-passing patch. The companion benchmark data is catalogued at Datasets/Benchmarks.md / SWE-bench; paired with Papers.md ref #155 (Jimenez et al. 2024). The standard tracker for whether a coding agent can be trusted with the bioinformatics-pipeline and analysis-code maintenance cell-ag teams increasingly delegate.