Data and licensing notice
MoleQuant screens candidates against open reference datasets licensed for commercial use, with attribution. This page states which, under what terms, and which dataset is deliberately not used.
Attribution
Contains data from the Materials Project (materialsproject.org), licensed CC BY 4.0. Additional reference data from PubChem and NIST. MoleQuant does not use non-commercially licensed datasets in its commercial service.
That statement appears in the footer of the MoleQuant website, and it is here in full because the licence requires attribution to be carried rather than implied.
What is used
- Materials Project
- Computed properties and crystal structures for inorganic materials. Licensed CC BY 4.0 — commercial use permitted, attribution required. materialsproject.org
- PubChem
- Chemical identity and property records, published by the US National Center for Biotechnology Information and placed in the public domain.
- NIST
- Reference physical and thermodynamic data published by the US National Institute of Standards and Technology.
Each record MoleQuant retrieves carries the database it came from, its version where one is published, and the date it was retrieved. That is not housekeeping: a figure whose provenance is unknown is indistinguishable from one that was invented, and a document generated next year has to be able to say what the source said when the comparison was made.
What is not used, and why
There is a larger and better-known materials dataset that MoleQuant does not use: Google DeepMind’s Graph Networks for Materials Exploration database. Its code is Apache 2.0, which is the distinction that makes this easy to get wrong, but its data is published under CC BY-NC 4.0 — NonCommercial.
MoleQuant has paying clients and a metered API. Using that data in a client deliverable would be commercial use, and that licence does not permit it. So it is not used, and no MoleQuant document, badge or figure is derived from it.
This matters commercially as well as legally. A larger database makes a better headline, and for a while this site carried one. But a smaller database we are entitled to use beats a larger one we are not — and a screening claim that cannot survive a licence question is worth nothing to the client relying on it.
The MoleQuant account with the Materials Project is registered as commercial, which means non-commercially licensed mirrors are excluded at the source rather than by a promise here.
Your data is not reference data
Nothing above applies to what you send us. Your challenge data and your results belong to you, each client’s data is isolated from every other client’s, and one client’s confidential data is never used for another.
There is also no training step. What MoleQuant retains is retrieval: a result is recorded, found again by relevance, and placed in the context of a later question — with its provenance attached, so a measured laboratory result and a predicted value are never treated as equivalent. No model is fine-tuned on your formulations.
Where a general insight is worth sharing across clients, it is reviewed and generalised first by a person, and the release is refused outright if the client’s formula, company or project name is still in the text. The original detailed record stays with the client who produced it.