Protein Structure and Function. Application of Bioinformatics Methods - John Rigden 2014

Prediction of protein function from its surface properties
Ligand-protein interaction
Prediction of drug sensitivity

The current diversity of disease-related Proteins that can successfully serve as targets for small drug molecules is restricted to a small number of Protein Families; due to sequence similarities, this number can be expanded to only 5% of all proteins in the human proteome (Hopkins and Groom 2002). The search for other potential drug candidates using experimental Methods, such as high-throughput Ligand screening, is time-consuming and expensive, while in 60% of new drug discovery projects, the target ultimately proves to be 'undruggable' (Brown and Superti-Furga 2003). Predicting whether an Active Site can bind drug-like molecules is one of the emerging Challenges in Protein structural biology (Cheng et al. 2007).

A recent study describes an approach that builds upon the aforementioned geometric pocket-finding algorithms by incorporating protein surface desolvation and surface curvature metrics (Cheng et al. 2007). These innovations were combined with evaluations of typical protein–ligand interaction features—such as the correlation between ligand molecular weight and buried protein surface area—into a single empirical score that assesses the potential for drug-like binding. A similar metric can be derived from binding pocket analyses performed using nuclear magnetic Resonance (Hajduk et al. 2005). Both approaches converge on the Conclusion that drug molecules and their analogs preferentially bind to large, easily desolvated pockets of complex geometry.