Will better knowledge of protein folding mean that we can predict rxns?

[Simfish]

These days, it seems that we all rely on trial and error in order to predict the biochemical reactions of new drugs (say, we want a drug to be a ligand, but we have to rely on trial and error to predict whether or not it will actually fit - plus - we also need trial and error to determine the binding affinity of the ligand and how the binding affinity of the ligand compares to that of other ligands).

And how does this also apply to toxicology? We have some theory to predict how toxic something *could* be. But our knowledge is still woefully incomplete, and no one's really convinced until we actually run clinical trials. Might knowledge of protein folding help with that? (of course, clinical trials will still always have to be run, but would they be significantly more convincing [and better targeted] if we had better knowledge of protein folding?

 

[Ygggdrasil]

Currently researchers designing new drugs can and do use computational methods to help improve those drugs. If a structure of the target protein is available, researchers can use computer programs to help determine where the drug is binding. The program will search along the surface of the protein for suitable drug binding sites and "dock" the molecule into the binding site. Seeing how and where the molecule binds can be helpful in determining how to change the structure of the drug to improve binding. Of course, these predictions are not always right and sometimes the results of these computational methods help us understand the results of trial-and-error drug screening rather than replace it.

There are, however, major limitations with these methods. First, these methods require a known structure for the drug target or at least a structure of a protein that is similar enough to the target. There are methods for predicting protein structure from the amino acid sequence of the protein, but I don't know if these predictions are reliable enough to generate good drug candidates in a computational screen. Second, our models of intermolecular interactions, while useable, probably aren't entirely correct and need more refinement. Finally, we have very little understanding of protein dynamics, that is, how proteins move around due to thermal motions and how this thermal "breathing" affects their function and drug binding.

Thus, better knowledge of protein folding can help these issues. Better methods to predict protein structure would allow these computational methods to be used on targets where no experimental structure has been determined (many membrane proteins are major drug targets and their structures are notoriously difficult to determine experimentally). Research into protein folding can also help refine our knowledge of intermolecular interactions and is intimately connected to the problem of understanding protein motions and dynamics.

In theory, computational methods could help with toxicology by determining which other proteins in the body a drug might bind. This would be very computationally intensive (there are A LOT of proteins in the body to consider). However, there is a large issue here. We don't know the function of a large number of the proteins and other biomolecules. Thus, if we see that the drug will bind to protein X as well as the target molecule, we wouldn't necessarily know whether this cross-reaction with protein X would be problematic. Similarly, while computational methods can help design drugs that bind to the target protein better, computational methods won't help identify new drug targets for new diseases.

 

[Simfish]

Thanks for the reply - it perfectly answered my question! :)