Could Someone Guide me on on Implementing Custom Featurizer for Molecular Dynamics Simulations

Hello there,

I am a research student working on a project that involves molecular dynamics simulations of small organic molecules. I have been using DeepChem for some time now and have found it incredibly useful for many aspects of my work. Although; I have reached a point where; I need some expert advice on implementing a custom featurizer tailored to my specific needs.

To predict the stability of small organic molecules in various solvents using machine learning models.
I have a dataset of around 10,000 molecules with their; respective stability measures in different solvents.

I want to incorporate specific features related to molecular dynamics; such as solvation free energy; specific interaction energies; and dynamic properties like diffusion coefficients.

I am unsure about the best practices for designing a custom featurizer that can handle these dynamic properties effectively. Should I extend an existing featurizer or build one from scratch?

How can I seamlessly integrate my custom featurizer with the rest of the DeepChem pipeline? Are there any examples or documentation that could guide me through this process?

What are the best practices for creating a custom featurizer in DeepChem, especially for dynamic properties?

Also; I have gone through this post; which definitely helped me out a lot.

Are there any existing tools or libraries within DeepChem or externally that can help in calculating and incorporating these molecular dynamics features?

Thank you in advance for your help and assistance.