I thought I would use this blog to summarise the recent Alchemistry Free Energy workshop in Göttingen, Germany. This event, organised by MPI BPC and BioExcel, brought together academics and industrialists who work with alchemical MM methods to calculate free energies. This was a very successful successor to a similar event organised two year ago in London and now looks to be repeated yearly, alternating between Europe and Boston.

The highlight of the event was the keynote speaker, William Jorgensen, whose work was the basis of most of the topics discussed at the workshop.  He made the very valid point that the errors given for many of these calculations are too good to be true, and the field needs to find the root of this remarkably low error. He suggested that new forcefields need to account for ligand behaviour both in the bound and unbound states to truly capture what is going on. Oh, and that sampling is still an issue. He, like many others during the workshop, also stressed the importance of checking your initial model. Are your histidine residues protonated? Are there waters there or not? Does your ligand parameterisation work?  

Schrödinger’s FEP+ package came out as a clear favourite tool within industry, allowing users to calculate ΔΔG changes between similar compounds. Several talks presented data showing successful use cases, though some pointed out there were still steps needed to ensure correct usage by the end user. It was also nice to see that groups were looking at implementing, and refining, the method in the open source GROMACS package. Making it a tool more accessible to academic groups who cannot afford a license. 

There were less discussions on absolute binding free energy calculations (ABFE, giving ΔG) with the majority of talks discussing the computationally cheaper relative binding free energy calcs (RBFE, giving  ΔΔG between two ligands). This makes sense, as this is a very powerful metric to use in ligand optimisation. Though it is important to stress that these results are still dependent on at least one good crystal structure, ligands that are FEP friendly and ideally no charge changes. 

I will end by expanding on charges, as these are still a big issue within the field. In short: when you alchemically remove or introduce a charge group, the net charge of your system will change and artefacts will start to convolute your data. Most people still adopt the correction proposed by Rocklin et al, but it was promising to see that several groups are looking at developing additional alchemical methods to better estimate the impact of charge changes. I think UCB’s policy of filtering out these ligands before testing is a good, albeit cheeky, workaround for now.