GDB-20 Database - Part2

About 

GDB-20 enumerates small organic molecules up to 20 atoms of C, N, O, S, F, Cl, Br, and I following simple chemical stability and synthetic feasibility rules.

By combining systematic graph enumeration with machine learning–based generation, we assembled 12,092,137,338 unique molecules as part of the newest GDB-20 database, representing a subset of an estimated 32 trillion possible molecules within this chemical space.

How to cite

To cite GDB-20, please reference:

Ye Buehler, Sacha Javor, and Jean-Louis Reymond. “Sampling a GDB-20 Database of 32 Trillion Drug-Like Molecules by Generative Artificial Intelligence.” ChemRxiv, 2026, https://doi.org/10.26434/chemrxiv.15000288/v1.

Dataset DOI:

https://doi.org/10.5281/zenodo.17368725.

https://doi.org/10.5281/zenodo.17375415.

Download

You can download the databases and subsets of it using the links provided. All the molecules are stored in canonized SMILES format and compressed as tar/gz archive (for Windows users: Download 7-zip to open archives).

GDB-20 (Part1)

GDB-20-Set (50 thousand)     GDB20.50000.smi.gz    470.05 KB

GDB-20-Set (50 million)    GDB20.50000000.smi.gz     484.05 MB

GDB-20-Set (HAC1-17)    GDB20.HAC1-17.tar.gz     2.84 GB

GDB-20-Set (HAC18)     GDB20.HAC18.tar.gz    13.73 GB

GDB-20-Set (HAC19)     GDB20.HAC19.tar.gz     17.85 GB

GDB-20 (Part2)

GDB-20-Set (HAC20)     GDB20.HAC20.tar.gz    35.30 GB

Terms and conditions: The GDB databases may be downloaded free of charge. In published research involving GDB, cite the appropriate references mentioned above. GDB must not be used as part of or in patents. GDB and large portions thereof must not be redistributed without the express written permission of Jean-Louis Reymond.