This online talk will be given by Márton Vass from Schrödinger.
Registration: free, required to get the Zoom link. Click Here to Register.
Title: Structure-based target enablement with IFD-MD and FEP+
Abstract: AlphaFold2 and other recently developed machine learning protein structure prediction tools have sparked a new interest in enabling structure-based drug discovery for targets with no available relevant structural information. At Schrödinger, we are working towards understanding the possible uses and limitations of ML predicted structures and methods of refining them for accurate physics-based modelling.
We have previously shown that raw AlphaFold2 structures can provide virtual screening enrichments on a par with those achieved using apo protein structures and that this performance can be improved to similar levels seen with relevant holo structures using tools like Induced Fit Docking with Molecular Dynamics (IFD-MD).
The use of AlphaFold2 models can be extended further to ligand optimisation by combining IFD-MD refinement with FEP+, to derive models that are capable of recapitulating known SAR and which, subsequently, can be used for prospective free-energy calculations.
In addition to its use in on-target modelling, we’ve extended the use of IFD-MD to off-target modelling, where the use of a new and faster ‘consensus IFD-MD’ approach allows us to run the workflow on a larger number of input protein conformations. This method appears to show promise for enabling structure-based ADMET modelling by targeting commonly encountered off-targets, such as CYPs, PXR and hERG.