PoseBusters inputs

[Bae-SungHan]

Hi.
I have some questions about input of PoseBusters benchmark.

1. In the previous issue I made (#210), one of your members answered that you used biological assembly, not asymmetric unit, for parsing input sequence for PoseBusters benchmark. I understand that AlphaFold3 used asymmetric units as input. Is there any biological reason for this?

2. I parsed input sequences from PoseBusters prediction you provided, but I found some difference between biological assembly sequences from RCSB PDB and parsed ones. Is there any other method you used for parsing biological assembly sequences?

Thank you

[GriffinYoung]

Just to follow up on this, it'd be helpful for reproducibility if you could include the chai input fasta files which were used to generate the PoseBusters results as Boltz does here for their evals dataset: https://github.com/jwohlwend/boltz/blob/main/docs/evaluation.md