gbff

Tools for Bayesian forcefield development

#Prerequisites

• Install redis with conda install redis
• Install celery with pip install celery
• Get FreeSolv database (git clone https://github.com/choderalab/FreeSolv.git)
• Using commit 3acd4f6a5f005b803fac024a3a87f64b51409e28
• set FREESOLV_PATH to location of database

#Preparing the database

• Run code/rebuild_freesolv
• Initial simulations can be run before the sampling loop to speed up debugging
• This is enabled by default
• code/prepare-database --types typefile --parameters parameterfile --output outputpickle

#Running GBFF (examples in scripts)

• Start redis with the command redis-server
• Set the environment variable CELERY_CONFIG to point to hydration_energies/config.yaml
• Edit config.yaml in hydration_energies so that both fields point to the redis server
• Start worker with the command celery -A hydration_energies worker -l info -c 1 --app=hydration_energies.app:app
• the c option allows you to choose the number of processes/worker
• Run parameterize-using-database.py

#Output

• Defaults to hdf5 backend
• outputs in /cbio/jclab/projects/pgrinaway/gbff/outputs.tar.gz compressed
• 300_adaptive_3gbmodel_largejoint_days.h5 is the most recent dataset