Protein selectivity analysis

Quantum Pharmaceuticals recently proposed a new method for toxicity prediction based on computation of small molecules affinity to about 500 human proteins. The analysis of binding profiles for about 1000 known pharmaceutical agents led to establishment of a relation between the toxicological properties of a molecule and its activity against the selected representatives of approximately 50 protein families. This activity profile was further used as a natural set of descriptors for various toxicological endpoints predictions, including human-MRDD, human-MRTD, human-TDLo, mouse-LDso (oral, intravenous, subcutaneous), rat-LDso (oral, intravenous, subcutaneous, intra-peritoneal), etc. ... 

The comparison with experimental data is somewhat complicated by the fact that the entry to the second pocket may be closed by a salt bridge which has to be broken to gain access to the pocket. Nevertheless, good agreement of the selectivity analysis with the selectivity profile of known inhibitors is found. In this case, the authors come to the conclusion, that taking protein flexibility into account with MOVE = 1 better reproduces the experimental data. 

The selectivity analysis for the SI pocket is complex as it is surrounded by a loop. Its length and amino acid composition differs between the individual MMPs, leading to different shapes and interaction patterns for this subsite. Here, computational techniques like GRID/CPCA are especially advantageous, as they allow an automated, unbiased view on the interactions and an abstraction from a discussion of differences in single amino acids. They address the sum of all interactions at once, and the distances in the score plot allow one to somewhat quantify the differences among the proteins. 

They can be used directly in protein families in order to perform selectivity analysis based on consensus principal component analysis methodology. This technique helps to identify regions in the protein space that are selective for one enzyme. The selective region may describe the selective pattern of the target proteins. Moreover, this information can be used to compare different models of the same enzyme. 

The present analysis might give rise to a somewhat pessimistic view of the effectiveness of protein secondary structure prediction algorithms. In fact, with the increasing number of proteins with known three-dimensional structure, constant re-evaluation of performance must take place in order to ascertain the validity of the methods. We note that the methods do not have the predictive power claimed by its authors when analyzed consistently using the 148 proteins selected in this study. Moreover, the situation is even worse for the Mathews correlation coefficient, which indicates that the predictions are poorly correlated with the actual structure. 

Table 2.1 Some popular tools for performing selectivity analysis using protein sequence...

---