This application is available on Streamlit

This is a really quick and dirty experiment for using lz4 and Tucan to find similar molecules.

The code of this is on my alter-github account: https://github.com/bjonnh-streamlit/mol_search

The dataset is from the LOTUS database and Wikidata.

• LZ4: https://github.com/lz4/lz4
• Tucan: https://github.com/TUCAN-nest/TUCAN

The idea to use Tucan and play with all that came from Adriano Rutz, we are both part of the team behind LOTUS.

It is using Streamlit for its web ui.

Rdkit is used for the molecule massaging.

See those other projects for more information on why I am trying that (and Daniel's work is what started me on that contraption):

• https://github.com/daenuprobst/molzip
• https://arxiv.org/abs/2212.09410

To reproduce:

• Install and run poetry update
• Run get_smiles_wikidata.py (takes 12s)
• Run generate_database.py (takes 1min on a Ryzen 7 3700X)
• Run streamlit run main.py (almost instant)

• Personal website: https://www.bjonnh.net
• You can find me on mastodon too(t): https://mastodon.social/@bjonnh

Your molecules are never stored, the only people that could eventually see them are the streamlit people. And I doubt they care about your molecules. But if your molecules are super secret, it is like your extremities, don't insert them in machines you don't know or understand.

Also, this is an experimental tool meant to test things, you're not supposed to rely on it for anything important, and it comes with no warranty or support whatsoever.

https://raw.githubusercontent.com/bjonnh-streamlit/mol_search/main/LICENSE