Request to implement dense MSA pairing for structure prediction

Hi, I noticed Boltz uses dense MSA pairing (https://blog.stackademic.com/boltz-1-open-source-biomolecular-interaction-modeling-with-python-and-r-code-tool-playground-77e1e660294b) instead of using colabfold's pairing. This helps in saving substantial time during MSA generation as there is no need to use colabfold's MSA pairing method.

Will be great if this can be implemented in Protenix as this will help in using unpaired MSAs directly in Protenix during large scale structure predictions of multimers.

Regards
Rakesh

Start by tracing Protenix's current ColabFold MSA pairing path and how unpaired MSAs enter multimer structure prediction. Compare the expected dense pairing approach with the linked Boltz description; done means Protenix can use unpaired MSAs directly for large-scale multimer predictions without the existing pairing step.

Written by the indexing model from the issue text.