Start benchmarking fragmenstein atom placement

[mwinokan]

@roncv once you have successfully run python -m bulkdock setup Flavi_NS5_RdRp with the target I put on your SSD you can proceed with benchmarking and developing the fragmenstein placement code in this repo.

To get started:

• Clone the RdRp_FFF repo that has a bunch of compound designs for an active project I'm working on
• Copy this file into your BulkDock/INPUTS directory
• Run python -m bulkdock.batch place Flavi_NS5_RdRp cavity_b_fstein_bulkdock_input.csv to run some fragmenstein placements

Then things to explore are:

• Create some sort of statistics output from fragmenstein_place, maybe a JSON file or pickled pandas DataFrame that includes runtime statistics, statistics on successes of the placement
• How reliable is Fragmenstein at solving bioisosteric merge poses (vs pure merges)?
• How reliable is RMSD as a metric for pose quality?
• Spread of conformers produced by fragmenstein when repeating the same job
• Explore minimising / protonating / modifying the reference protein structures and see the effect on the placements
• CLI arguments to change some of the Fragmenstein parameters such as the minimiser engine (Victor vs Wictor in Fragmenstein lingo)
• What was Steph's quilter doing?

Obviously we'll need to meet to get into the details of this but maybe you can mull on some of these ideas while you are bored on the chair lifts