Given the ease of doing x-ray crystallography & solving complexed structures and given that it gives a snap-shot of ligand protein interactions, do you think that indirect deductions through QSAR analysis might still hold some promise in drug discovery?
Hi, I believe that QSAR analysis is still useful in modern drug discovery. In particular the use of 3D-QSAR model for fast screening in order to narrow down the number of compounds coupled to docking studies for evaluating/identifying novel scaffolds for a given target is still a valuable approach in med chem.
Yes, because x-ray crystallography results provide only a frozen view. Regards, Juan.
Although I am primarily working with structure-based methods, I'd definitely think that QSAR and similar methods are still of value.
To start with, not all targets can be readily crystallized. Membrane and trans-membrane proteins are for instance more difficult to crystallize. As a result not all targets can be addressed through crystallography. Homology modeling can obviously help, but the technique has its own pitfalls and not all homology models will be of sufficient quality.
The point that X-ray is a frozen view is very relevant. Or, actually, that one X-ray structure is one frozen view. Induced fit effects can sometimes be perceived / addressed through multiple structures. But if the structural information is limited, a docking approach might fail to recognize some ligands that require significant conformation changes in the protein, while a QSAR model will take it into consideration... as long as the relevant ligands are included in the training set!
In virtual screening, it is common to use a funnel approach whereby the least computationally expensive operations (substructure filters, similarity metrics or other) are performed first. If you have a good QSAR model that is faster to apply than docking, it can be used as a filter so to select the best candidates for the more demanding docking step.
Finally, next to QSAR, there's QSPR, where you're looking at properties (permeability, BBB permeation, stability, solubility....) rather than activity. Those models are also of interest for drug design and discovery, and cannot be replaced by crystallography.
QSAR is indeed less and less relevant in the context of an ongoing lead optimisation program. The QSAR modelers are often a step behind the medicinal chemists. They need time to collect and analyse relevant data, while the chemists often infer the right trends by themselves. In structure-based design, the data are a useful, but not critical component of the design. These days x-ray or NMR structures are available for the vast majority of therapeutically relevant targets (or their close homologues).
Also, the modeler's services are needed most at the early stage of the project, when there is no data to start a QSAR study.
I disagree with Dimitri's comment. Almost all medicinal chemists employ a reasoning based in Ancient Organic Chemistry (up to 1955, resonance, inductive effects, etc.). Quantum-chemistry based QSAR explains clearly the experimental resuts (binding affinities, anticancer, antiviral activities, etc.) provided by the medicinal chemists. Regards.