severe clashes - a note for future improvements

[Jameel9]

Hi Boltz team,

Thanks for making this software publicly available. I did gave it a try (as well as AF3 server) using the same challenging task (PDB ID: 6WW6) and Boltz flipped one base of the RNA as in the solved structure! but AF3 flipped three!
However, clashes were seen in both outputs (in Boltz clashes were severe).

AF3 (correctly flipped in grey color, wrong flipping in cyan, clashing in yellow)

Boltz (correctly flipped in grey color, clashing in yellow)

[gcorso]

Hi @Jameel9 thank you for the note! Indeed we are actively working on fixing these type of issues. We will hopefully have a model update solving these in the coming weeks, but in the meantime we recommend using some short physics-based relaxations if needed using tools like Amber (free) or PyRosetta (requires license)