This is what bothers me. An empirically derived weighting function. To me, ligand efficiency makes sense from a physical standpoint. It's a value that measure how well every atom contributes overall to the binding. It makes sense to me that bigger molecules are worse at using their atoms. This scaling function makes no physical sense to me. Somebody please explain to me (and admittedly I maybe totally miss the point), but please explain why Fit Quality is a better measure than LE from a physical standpoint of the origin of the measures.
Thank you for pointing your readers to our blog post Teddy. I had the same question when I read the original Reynolds paper on LE and FQ (Reynolds et al. Bioorg Med Chem Lett 2007). It does make sense that LE falls off with size, but from what I gathered the authors wanted a metric that would determine whether a molecule is as ligand efficient as it can theoretically be based on its size. I have a couple problems with this. First, the FQ parameters are based on empirical studies and as such are limited by the amount of data currently available. More importantly though, the metric doesn't take into consideration properties of the protein target itself, such as SASA of the binding cavity, etc., which surely play a role in this. Perhaps target specific indices would make more sense? Or maybe LE calculations should be restricted to fragment-sized moieties only?
ReplyDeleteI agree that there is something appealing about the simple, physically grounded definition of ligand efficiency (binding energy per heavy atom). Using empirical data in a fitting function is harder to understand on an intuitive level.
ReplyDeleteAs to the question of why larger molecules tend to have lower ligand efficiencies than smaller ones, there are a number of reasonable proposals, but one hypothesis I haven’t seen is that there is usually no need for affinities better than low nanomolar, either in nature or in drug discovery. And since it is generally possible to reach this level of affinity with molecules of reasonable size, there is no selection pressure (again, either Darwinian or human) to achieve higher ligand efficiencies. If this is the case, using an empirical fitting function would set the bar too low: one could do better, it’s just that it’s normally not worth the effort.
Still, I’m willing to be convinced. Has anyone used fit quality in a medicinal chemistry program and found it helpful?