AWE currently requires a legacy build of MSMBuilder to perform clustering per the tutorial. This version of MSMBuilder is not Python 3-compatible and several features used to extract models for the Walkers have been removed in newer versions. Additionally, MSMBuilder does some file format conversions that can lead to bad trajectories when extracted due to interpolating hydrogen positions.
The clustering process needs to be updated to use a more recent version of MSMBuilder.
MSMBuilder uses scikit-learn to do clustering. If converting to a newer version of MSMBuilder is not viable, AWE should include its own clustering utility built on scikit-learn to simplify the installation process and system preparation.
jeffkinnison
commented
8 years ago
AWE currently requires a legacy build of MSMBuilder to perform clustering per the tutorial. This version of MSMBuilder is not Python 3-compatible and several features used to extract models for the Walkers have been removed in newer versions. Additionally, MSMBuilder does some file format conversions that can lead to bad trajectories when extracted due to interpolating hydrogen positions.
The clustering process needs to be updated to use a more recent version of MSMBuilder.
MSMBuilder uses scikit-learn to do clustering. If converting to a newer version of MSMBuilder is not viable, AWE should include its own clustering utility built on scikit-learn to simplify the installation process and system preparation.