concept · intermediate · partly written
Reaction data: what is public, what is proprietary, and what each holds
Public reaction data are mostly text-mined patents and curated enzyme reactions; the larger, richer sets are proprietary.
Also called: reaction databases, patent reaction sets, USPTO reaction dataset, Open Reaction Database, Rhea, BRENDA
Which reaction data can a biologist actually reuse?
Public: patent-mined reactions (USPTO, CC0 per a blog), the Open Reaction Database (CC-BY-SA), Rhea (CC BY 4.0), BRENDA (free for academics). Proprietary: Reaxys, Pistachio and company notebooks. Each public set has known errors, and only pieces of them have been measured. 12345
Reported, not yet replicated
What is in the public USPTO reaction set, and how clean is it? more
The set holds 3,305,795 reactions from US patents 1976 to 2016, which the blog that announced it calls CC0. Another count puts the raw number at 3,748,191. A model-builder's study found only about 2% of reaction templates are shared by all four of USPTO, Reaxys, Pistachio and an AstraZeneca notebook subset. Whole-set error was not published; a 100-reaction check of the original miner exists, and one group estimates from a 50-entry sample that as much as half of USPTO entries may be compromised against about 10% for Reaxys. That estimate comes from a single group. 1567
Reported, not yet replicated
What do the proprietary reaction sets hold, and what do we know about them? more
Reaxys held 6,540,786 reactions in a 2020 template study, with conditions, solvent, catalyst and yield. Pistachio held 6,836,027 in November 2017, and an AstraZeneca notebook subset held 398,779. A group at Cambridge analysed 15.4 million filtered Reaxys reactions and found 54% had no yield, 53.9% no temperature and 98.4% no pressure. That sample was filtered from 33.5 million downloaded records, so the percentages do not describe all records. Terms of use were not read. 58
Reported, not yet replicated
Which open resources hold enzyme reactions and kinetics? more
Rhea held 13,673 reactions in release 119 (2021) under CC BY 4.0 and is the reaction vocabulary UniProt uses. BRENDA held about 332,000 annotated kinetic points in 2017, free for academic use. The Open Reaction Database stores structured conditions and outcomes, with data under CC-BY-SA and code under Apache, and added about 39,000 formerly proprietary high-throughput reactions in July 2024. A STRENDA audit found all 11 sampled kinetics papers lacked a needed item. EnzymeMap authors report mistakes inside BRENDA, including missing stereochemistry and unbalanced reactions. The size claims for Rhea and BRENDA are each database's own. 3429101112
Reported, not yet replicated
His hypothesis: there may be analogies in biology more
Amit's words: "Data gathered from food companies doing food manufacturing: they've done a lot of research into compounds, what reactions work, and how those things. There may be analogies in biology." This is his hypothesis, kept in his words. The sourced account sits in the next section. 13
Amit's hypothesis
What do the sources say about that hypothesis? more
Sources opened show chemistry's methods being reused on enzymes and metabolism, with measured gains. They show no process-engineering or food-company reaction dataset reused for a biological model. The search was shallow and company holdings are closed, so this is an absence in what could be opened, not proof of absence. 141516
Our inference
Plain answer
Why it matters
A model or a knowledge graph that wants reaction facts can only use what it may reuse, and each resource carries different errors and gaps. 5
Our inference
Real numbers
USPTO patent reactions
3,305,795 reactions from US patents 1976 to 2016, described by a vendor-side blog as CC0. 1
Reported, not yet replicated
Reaxys in a 2020 template study
6,540,786 reactions after an initial internal curation step. 5
Reported, not yet replicated
Pistachio (November 2017)
6,836,027 reactions; an AstraZeneca notebook subset held 398,779. 5
Reported, not yet replicated
Rhea release 119 (2021)
13,673 reactions, on ChEBI participants, with EC and UniProt links. 3
Reported, not yet replicated
EnzymeMap
63,137 atom-mapped enzymatic reactions within their EC class; 8,599 unbalanced and 5,841 unmappable balanced BRENDA entries were found. 11
Reported, not yet replicated
Still open here
- How large is the Open Reaction Database now? Only a July 2024 addition of about 39,000 reactions was found.
- What is the size of SABIO-RK? Not verified.
- What is the overall published error rate of the public USPTO set? No whole-set audit was found.
Not yet written: Equation.
Linked from
- Chemistry-trained reaction models applied to enzymes and metabolism: what was measured
- Food composition databases are not reaction databases: what public food data hold
- Metabolite damage and repair: spontaneous chemistry inside cells that reaction resources mostly omit
- How a reaction gets from a patent paragraph into a database record, and what goes wrong
- How a reaction is written for a computer: reaction SMILES, atom maps and templates
- Open Reaction Database
- Jean-Loup Faulon
Sources
- The NextMove patent reaction dataset. Depth-First (blog) (2019). web, not peer reviewed; opened 2026-11-17 all records citing this
- Open Reaction Database documentation: overview. docs.open-reaction-database.org (2026). organisation-page, not peer reviewed; opened 2026-11-17 all records citing this
- Parit Bansal, et al.. Rhea, the reaction knowledgebase in 2022. Nucleic Acids Research (2022). paper; opened 2026-11-17 all records citing this
- Sandra Placzek, et al.. BRENDA in 2017: new perspectives and new tools in BRENDA. Nucleic Acids Research (2017). paper; opened 2026-11-17 all records citing this
- Amol Thakkar, et al.. Datasets and their influence on the development of computer assisted synthesis planning tools in the pharmaceutical domain. Chemical Science (2020). paper; opened 2026-11-17 all records citing this
- Daniel Lowe. Extraction of chemical structures and reactions from the literature. University of Cambridge (PhD thesis) (2012). paper, not peer reviewed; opened 2026-11-17 all records citing this
- Beker et al.. Machine learning may sometimes simply capture literature popularity trends [title not copied from the page]. journal not copied (2022). paper; opened 2026-11-17 all records citing this
- Jacob et al.. Analysis of Reaxys reaction records for missing conditions [title not copied from the page]. Journal of Cheminformatics (2017). paper; opened 2026-11-17 all records citing this
- Open Reaction Database update. macinchem.org (blog) (2024). web, not peer reviewed; opened 2026-11-17 all records citing this
- STRENDA authors. Audit of reported enzyme-kinetics data against the STRENDA guideline [title not copied from the page]. journal not copied (2018). paper; opened 2026-11-17 all records citing this
- Esther Heid, et al.. EnzymeMap: curation, validation and data-driven prediction of enzymatic reactions. Chemical Science (2023). paper; opened 2026-11-17 all records citing this
- Steven M. Kearnes, et al.. The Open Reaction Database. Journal of the American Chemical Society (2021). paper; opened 2026-11-17 all records citing this
- Amit's dictation of 2026-11-18: how biology data is represented, and where ontologies can be innovated. Kept in the atlas repository; not published.
- David Kreutter, Philippe Schwaller, Jean-Louis Reymond. Predicting enzymatic reactions with a molecular transformer. Chemical Science (2021). paper; opened 2026-11-17 all records citing this
- FoodData Central. U.S. Department of Agriculture (2026). database, not peer reviewed; opened 2026-11-17 all records citing this
Where it sits on the map
History of this record (1)
This record has been changed 1 time, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.
- Status
- partly written
- Last updated
- 2026-11-17
- Record
- content/concepts/reaction-data-resources-public-and-proprietary.json