How to read the output of flex ddG?

[Sh-Yana]

Dear Rosetta community,

I have a question on how to read the output of the “flex ddG” protocol – the output table is huge, I am new to both flex ddG and Rosetta, and would be eternally grateful if anyone could provide some clarity for me :)

The protocol predicts changes in binding free energies upon mutation in a protein-protein interface (interface ddG). I’ve run it with the recommended input parameters:
nstruct = 35
max_minimization_iter = 5000
abs_score_convergence_thresh = 1.0
number_backrub_trials = 35000.
As suggested by the “flex ddG” authors, the backrub_trajectory_stride parameter was set to 35000 (equal to the number_backrub_trials value) to minimize running time and disc usage.

The output for one mutation is attached as a screenshot-1 of the output-results.csv file, where some columns and rows were hidden to declutter the view, as the table is large. To know how to read the output, I would need to understand the following points:

1. what number in this table is the final result – the interface ddG I need? Do I understand it right that, in my example, the 0.465144337644352 value in the F9 cell is this result? That it is the ddG (the difference between the mutant and WT), calculated using the fa_talaris2014 function and re-weighted with the GAM model after 35000 backrub sampling steps, and averaged across the 35 structural models? In other words, is averaging across structural models already implemented in the protocol?

2. Then, what are the values in the cells F6-8 (for nstruct 10, 20, and 30, respectively) needed for? Were these values averaged across 10, 20, and 30 structures to check when the ddG calculation reaches a plateau, and to estimate whether the nstruct parameter can be adjusted (e.g. lowered to reduce the calculation time; or increased if the plateau was not reached, meaning that the sampling was not sufficient)? Is my understanding even remotely correct?

3. If I set the backrub_trajectory_stride parameter to a smaller value – let’s say, 1000, – this will produce 35 checkpoints (results from the earlier time points in the backrub trajectory) in the output file (kindly see the screenshot-2). Does it mean that the final ddG (after 35000 backrub steps, averaged across 35 structural models) can be found in the cell F17?

Thank you so much in advance!
Yana