The availability of a protein structure is common in drug discovery projects. This has been driven by the rapid expansion and availability of experimentally derived structures, as both publicly available structures that can be downloaded from the Protein Data Bank (PDB) and as proprietary solved protein structures developed within a discovery project. In addition to protein crystal structures, CryoEM structures give us glimpses into the world of molecular machines, which offer insights of how proteins function in combination.
With these available protein structures, it is convenient to accept the structures as given; however, every structure should be interrogated with respect to its specific context. The availability of protein structures offers an immensely powerful tool in drug design; but for a protein structure to inform a project, it first needs to be evaluated and protein preparation carried out to take account of aspects such as:
- Is this protein structure relevant to my project? A ligand bound structure should contain a ligand of equivalent size and shape to that within the ligand data set for the project.
- Is the protein in the correct state? There may be factors driving the structure into an alternative conformation or protonation state.
- Is the protein structure biologically relevant? The protein should represent the biological state of the protein, with the structure being well resolved in areas important for ligand binding.
The large amount of data available has changed the face of drug design, now opening the opportunity to train AI models. This provides insights such as observation of molecular machines in action, identification of unexpected protein interactions, and greater understanding of protein function and cooperability. We have a wealth of structural data available for us to apply to a project, but it is important to take time to fully understand the system under investigation.
- Nature 1958 Mar 8;181(4610):662-6. doi: 10.1038/181662a0
- Protein Data Bank, https://doi.org/10.2210/pdb1MBN/pdb
