[Question] Evaluating Amber ff14SB and espaloma on Pepconf

[LeifSeute]

Hello,

I am trying to reproduce the QM-fitting results from the table in figure 4a) from the paper (https://arxiv.org/abs/2010.01196). There, you compare espaloma with Amber ff14SB as baseline on the dataset that you created by sampling from optimization trajectories of pepconf (https://github.com/aoterodelaroza/pepconf).

I am wondering how you applied Amber ff14SB to the dataset since, to my knowledge, this force field requires information on the protein residues of the individual atoms. These can neither be found in the "pepconf"-dataset that is provided in the QM fitting tutorial (https://espaloma.wangyq.net/experiments/qm_fitting.html) nor in the one on qcarchive.

Also, the pepconf dataset on qcarchive is currently incomplete since a majority of the trajectories there cannot be loaded due to convergence issues (openforcefield/qca-dataset-submission#322) and the one from the tutorial mentioned above has a different number of molecules/snapshots than given in figure 4a) of the paper.

Any help on the dataset and obtaining the Amber ff14SB energies for reproducing the pepconf results of the paper would be appreciated!