search

choderalab / feflow

Recipes and protocols for molecular free energy calculations using the openmmtools/perses and Open Free Energy toolkits
MIT License
10 stars 1 forks source link

HTF migration #24

Closed ijpulidos closed 4 months ago

ijpulidos commented 5 months ago

Changes to adopt the HybridTopologyFactory class/object from openfe.

~The idea is migrating the class and also adapt it to meet the work from Fleck et al~

These changes are a starting point where we can test our implementation of HTF, at least minimally to what perses was testing, and gradually improving upon that.

ijpulidos commented 5 months ago

This is still in a very WIP status. I plan to work on this in the following days.

Currently I created a hybrid_topology.py module in utils, but I'm not sure that's the best place to put it. We could probably think about having an alchemy module or relative or something similar.

Currently I copied/pasted the test_relative.py module that perses uses to test the HTF. We would need to drop all the geometry engine and topology proposal things. And adapt the current code to use the feflow object itself.

ijpulidos commented 4 months ago

Temporarily we will be relying on perses to test many aspects of the HTF. Eventually we want to move away from depending on perses for the tests. Issues will be raised to follow up on this.

I tried organizing/modularizing things a bit, but not sure if this is what we want. I'm imagining this as having one module for more integration/science tests in test_relative.py and another more for the actual code unit tests for the HTF class in test_hybrid_topology.py, if that makes any sense. Thoughts?

For now the LambdaProtocol tests are failing since our implementation doesn't add the default functions automatically as it was done in perses. Do we want to leave it like this?