CAMELOT.py is a Python package that automates the process of learning bonded and non-bonded interaction parameters for coarse-grained simulations.
CAMELOT.py was written by Alex Holehouse during his time in the Pappu lab and is based on ideas developed by Dr. Kiersten Ruff which were applied to systematically coarse-grain various systems in the original CAMELOT manuscript [1]. This version represents a Python re-write of the original code, and we plan to re-structure this code into a standard Python package.
In particular, CAMELOT provides a general, reference framework for applying GPBO to learn parameters for coarse-grained simulations and can be integrated to work
Specifically, CAMELOT.py contains two independent stages.
-
Stage 1 analyzes all-atom simulations and extracts out bonded-terms using an inverse-Boltzmann approach to parameterize bond lengths, angles, and dihedrals. For more details on this see the accompanying manuscript by Ruff et al. [1].
-
Stage 2 integrates Gaussian Process Bayesian Optimization into the procedure of learning non-bonded interactions. The code utilizes the Gaussian Process for Machine Learning software package, although this is opaque to the users as CAMELOT.py autogenerates all of the underlying code needed to interact with and run the associated parameterization.
CAMELOT.py is fully developed in terms of the underlying models, although remains less user-friendly than it could/should be. We are working on this - if you need access to CAMELOT.py now the safest thing is to contact Alex directly (alex dot holehouse at wustl dot edu).
[1] Ruff, K. M., Harmon, T. S. & Pappu, R. V. CAMELOT: A machine learning approach for coarse-grained simulations of aggregation of block-copolymeric protein sequences. J. Chem. Phys. 143, 243123 (2015).