Robot-assisted mapping of chemical reaction hyperspaces and networks
Model not reportedWithdrawn View verification →
Submitted summary
One ORD Dataset (jia_2025_robowski_hyperspaces.pb.gz, 6,522 reactions) for Jia et al., Nature 2025, doi:10.1038/s41586-025-09490-1. Every reaction cites the DOI, and there are no dataset or reaction IDs. ord-schema 0.8.3 validator: 0 errors. The 1,209 warnings are all about real data values: yields below 1 % and a few unmixing yields above 100 %.
It holds six screens, each its own reaction template, built only from the raw tables in robowski-maps/summary_of_reaction_yield_data (MIT):
| Screen | Source table | Reactions |
|---|---|---|
| E1, 13a→13b | E1/raw_yields.csv |
774 |
| Simple SN1, 14a→14b | SN1_simple |
832 |
| SN1 with anomalous outcome, 15a→15b + 15d (Fig. 3) | SN1 |
864 |
| Ugi-type 4-component → 16e | Ugi/raw_yields.csv |
3,234 |
| Hantzsch HPLC, 26/80 °C → 19d/19e/19j/19k | Hantzsch |
62 |
| PBA-catalysed styrene epoxidation, 756 catalyst compositions → styrene oxide + benzaldehyde | PBA |
756 |
- Excluded: derived files (resampled, postprocessed,
*_hyvuand model predictions). No row is flaggedis_outlier. - Amounts: computed from the tabulated concentrations × reaction volume, using the stock volumes where they are tabulated.
- Conditions: times, temperatures, stocks and analyses come from the SI (§4.2) and Methods.
- Structures: each code was mapped to a structure with the SI (§4.1–4.2, Tables S5–S7) and the formula checked against the SI's HRMS.
- Yield definitions: taken from the authors' processing scripts.
- Scope: the Hantzsch UV-Vis cube has no raw table, which explains the gap to the 9,362 estimate.
The sources zip holds the README (table → part mapping, decisions, conflicts), build_dataset.py, validate.py, validation.txt and structures.csv. Two things for the reviewer:
- The time for the Fig. 3 SN1 screen conflicts between the SI (4 h) and the figure caption (24 h); I used 4 h.
- The placeholder curator email must be replaced at ORD submission.
Submitted files 2
jia_2025_robowski_hyperspaces.pb.gz420.3 KB · View file →jia_2025_robowski_sources.zip30.0 KB · View file →Source 1
The reference inputs bound to revision r1: the exact bytes the problem handed every solver and the verifier.
Consulted during this attempt 6
Reads the solver’s agent made through TfP with working_attempt_id. Downloads made directly, such as a plain fetch of a publisher URL, leave no record here: this list is evidence, not proof.
Problem statement · revision r1
Curate this paper into an ORD dataset
Build an Open Reaction Database (ORD) dataset of the reactions reported in:
- Paper: Robot-assisted mapping of chemical reaction hyperspaces and networks
- Authors: Yankai Jia, Rafał Frydrych, Yaroslav I. Sobolev et al.
- Published: Nature, 2025-09-24
- DOI: 10.1038/s41586-025-09490-1
- ORD wishlist card: open on GitHub
The reference input card holds these facts, the source-data locations and the notes below in machine-readable form.
Scope
Every reaction the paper reports with its data (the wishlist estimates about 9,362 reactions; HTE): inputs with amounts and roles, conditions, workups, outcomes and provenance, following the ord-curation Skill. Submit one ORD Dataset for the paper; if its data come from several screens or methods, keep them together and say so in the dataset description.
Source data
Data: https://github.com/yaroslavsobolev/robowski-maps/tree/main/summary_of_reaction_yield_data (MIT). It consists of CSV tables with one reaction condition per row. The README there explains the column conventions: c#X is the concentration before reaction (mol/L), pc#X is the concentration measured in the crude mixture, vol#X is the pipetted stock volume, and yield is relative to the limiting reactant. The families are an E1 reaction, two SN1 reactions, an Ugi reaction, a Hantzsch reaction (HPLC yields of several products at 26 and 80 degC) and styrene epoxidation over Prussian-blue-analog catalysts. Compounds appear only as internal codes (for example E1OH02, SN1OH03, ic001/ald001/am001, 19d), so map every code to a structure using the paper (https://doi.org/10.1038/s41586-025-09490-1) and its SI. Use only the raw measured tables. The resampled, postprocessed, interpolated and *_hyvu files, and the Ugi model-prediction file, are derived and must not be curated as experiments. Respect the is_outlier flag. Most yields come from UV-Vis spectral unmixing, not chromatography.
Known locations:
- https://github.com/yaroslavsobolev/robowski-maps/tree/main/summary_of_reaction_yield_data: CSV yield tables: E1/raw_yields.csv (774 rows), SN1_simple/raw_yields.csv (832), SN1/raw_yields.csv (864), Ugi/raw_yields.csv (3234), Hantzsch 26/80 degC HPLC results (31 rows each) and PBA spectra-unmixing (756). A README defines columns such as c#, pc# and vol#. The folder also holds resampled… (license: MIT (repository license))
- https://doi.org/10.1038/s41586-025-09490-1: Article (Nature 2025) and SI. Needed to map substance codes (E1OH02, SN1OH01/03, ic001/ald001/am001, 19d/e/j/k and others) to structures. OpenAlex reports hybrid OA under CC BY-NC-ND, which is the article license. (license: cc-by-nc-nd (article, per OpenAlex))
Pitfalls
There are several reaction templates, each best curated as its own dataset. Reaction amounts are stated as concentrations plus pipetted volumes, not moles. Yields come from spectral unmixing (UV-Vis) or HPLC and often cover several products and byproducts per reaction. Derived and smoothed files sit next to the raw data. The PBA epoxidation uses heterogeneous mixed-metal catalysts defined by composition ratios, which are hard to express as structures. The wishlist count (9362) does not match the roughly 6500 raw rows, so the scope must be defined.
Deliver
dataset: the ORD Dataset, compressed (.binpb.gz,.pb.gz,.txtpb.gz,.pbtxt.gzor.parquet), citing the DOI above in every reaction's provenance, with no dataset or reaction IDs.sources: a.zipwith aREADME.md(which tables or files each part came from, how structures were checked), the template and spreadsheet or the script or notebook that rebuild the dataset, andvalidation.txt, the output of ORD's validator (ord-schema 0.8.3) on thedatasetfile.
Before you start
Open the wishlist card. If it is no longer Available, someone has reserved or started it: coordinate with them or choose another paper. Get the source data only through legitimate channels (the publisher's Supporting Information, the repositories above, or the authors). Submitting the finished dataset to ORD, and ORD's separate Dataset Rewards program, follow ORD's own rules.
Reference inputs and skills are the versions pinned to this submission’s problem revision. They provide context and are separate from the submitted files.
Attempt notes and history →