While there currently are efforts in the works to extend the RosettaLigandEnsemble approach (https://pubmed.ncbi.nlm.nih.gov/29732444/) to ensembles of proteins and not just ensembles of ligands, right now that’s still in the development phase, and has not been released.
Right now the best suggestion would be to run separate docking runs on each individual structure. (After reducing the total number of structures by clustering or the like.) DeLuca, Khar and Meiler (https://pubmed.ncbi.nlm.nih.gov/26207742/) have suggestions for reducing the computational cost for screening multiple ligands. Those suggestions can also be implemented to reduce the computational costs of screeing a single ligand against mulitiple protein structures.
.
