Target protein with calcium ions

[daizc-fdu]

Hello, what a great software! Thank you for developing it!
I am designing a binder for a protein named ANXA11 with calcium ions. The calcium ions form ionic bonds with several loops of ANXA11. I would like to know if ODesign can solve the design of binders for such target (ANXA11) with non-covalent binding(Ca2+), and what adjustments should be made to the json file?

[daizc-fdu]

I have already used some other protein design algorithms, but during the refolding step before scoring, AF3 considers that the binding of Ca2+ to the binder is more conducive to the stability of the system, thereby the binder directly "snatching" the Ca2+ from ANXA11. For example, as shown in the figure below, this will make the subsequent scoring completely unreliable. Because essentially, it is not generating a binder to disrupt the function of the (ANXA11 + calcium ion) system, but rather creating a calcium ion chelator, which deviates from our design purpose.
In the picture below, the purple one is the binder, which "takes away" the calcium ions that should be on ANXA11.

[OdinZhang]

That is a very good question. I think this is one of the major limitations of AF3: in some cases, it refolds otherwise correct designs into incorrect conformations.

Regarding your question, one possible approach is to treat Ca²⁺ as a ligand and design binders accordingly. However, to be frank, I do not expect this strategy to work very well, as this scenario is not a classical setting covered during model training.

A more principled solution would be to train a dedicated model with an ion + target → binder formulation. Alternatively, if you are interested in collaboration, we would be happy to discuss arranging a project focused on this direction.