Check it out, and feel free to drop your questions here or on GitHub!
🔗 GitHub: github.com/THGLab/HiQBind
📄 Paper: pubs.rsc.org/en/content/a...
#machinelearning #proteinligand #proteinligandbindingaffinity #structuralbiology #ai4sci
Huge shoutout to the amazing team behind this:
👏 Lead author Yingze (Eric) Wang and @kunyangsun.bsky.social
👏 PI Prof. Teresa Head-Gordon
👏 Teammates Jie Li, Xingyi Guan, Oufan Zhang, Dorian Bagni
👏 Collaborators Dr. Heather A. Carlson and Prof. Yang Zhang
What’s next?
We’re exploring:
🔁 Rotamer refinement
🤖 Binding label extraction with LLMs (maybe 👀)
🧠 Better data splits (possibly inspired by PLINDER) to support ML research!
Since we're focused on structural data with binding labels, we applied this workflow to major open-access datasets (BioLiP, BindingDB, and BindingMOAD) to generate HiQBind: a cleaned, corrected dataset comparable in size to PDBBind v2020 but with significantly improved structural quality! 💥
In this work, we built HiQBind-workflow, a semi-automated workflow that processes protein–ligand structures from the RCSB PDB by adding missing atoms, correcting ligand geometries, fixing bond orders and protonation states, and much more!
Copying Oliver's post from Linkedin to help us gain some visibility here!
🚨 Our paper is out! 🚨
"A workflow to create a high-quality protein–ligand binding dataset for training, validation, and prediction tasks" is now published in Digital Discovery! 🎉
We have left X for greener pastures and bluer skies - good riddance to Nazi-saluting Musk and his abuse of science and engineering that in fact enriched him.