`multi_pose_sampling` does not write `summary.csv` and does return per residue pLDDT scores

[jsture]

summary.csv is only generated by benchmarking functions and not the main predictor function. In addition, per residue scores are not returned.

One would probably have to add:

if confidence:
    plddt, plddt_lig = model.run_confidence_estimation(
        sample, output_struct, return_avg_stats=True
    )

    sample = model.run_confidence_estimation(
        sample, output_struct, return_avg_stats=False
    )

Then append that to some more lists and write those two as .csv files.

The logic and output of run_confidence_estimation is a bit too non-straightforward and undocumented for me to add this myself.

[Samuel-gwb]

Yes, it works fine just printing plddt and plddt_lig, while "sample" gives huge matrix setting return_avg_stats=False.

[amorehead]

I believe this PR addresses the missing plDDT values issue.