I am trying to create a ML model which includes various energy terms accounting mainly for the enthalpic contribution to ligand-binding. However, entropic contribution is neglected, although it plays a very important role, thus I would like to include it implicitly by adding QSAR descriptors related to ligand flexibility. So far I could think of these descriptors:
number of rotatable bonds,
number of heavy atoms
Molecular Weight/number of atoms
Which other QSAR descriptors do you believe could help to quantify the change in ligand entropy upon binding?
In this reference is discussed.
Wei, D.T.; Meadows, J.C.; Kellogg, G.E. Effects of Entropy on QSAR Equations for HIV-1 Protease: 1. Using Hydropathic Binding Descriptors. 2. Unrestrained Complex Structure Optimizations. Med. Chem. Res. 1997, 7, 259-270
The number of non-hydrogen atoms tends to correlate with the non-electrostatic interaction energy contribution to the binding free energy:
Article Aflatoxin B1–Formamidopyrimidine DNA Adducts: Relationships ...
You might be interested by this paper:
Beyond Rotatable Bond Counts: Capturing 3D Conformational Flexibility in a Single Descriptor
J Chem Inf Model. 2016 56, 2347–2352.
Robert Paton I was aware if this paper. In fact, I implemented the descriptor in Python and it was bad. Unlike what they say, RDKit's conformer generator and the MMFF force field are no match for Schrodinger's confgen and OPLS3! The simple rotatable bond count is much better than that descriptor for both classification (actives/inactives) and regression (affinity ranking) problems.
Martiniano Bello the paper you cited is ancient, I couldn't even find it in pdf.
To answer my question, I cite below a new paper that describes exactly what I had in mind!
https://chemrxiv.org/articles/preprint/Understanding_Conformational_Entropy_in_Small_Molecules/12671027/1
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Atomic, Molecular and Optical Physics
- Atoms