Divide and Conquer Strategies for Protein Structure Prediction

In this chapter, we discuss some approaches to the problem of protein structure prediction by addressing “simpler” sub-problems. The rationale behind this strategy is to develop methods for predicting some interesting structural characteristics of the protein, which can be useful per se and, at the same time, can be of help in solving the main problem. In particular, we discuss the problem of predicting the protein secondary structure, which is at the moment one of the most successful sub-problems addressed in computational biology. Available secondary structure predictors are very reliable and can be routinely used for annotating new genomes or as input for other more complex prediction tasks, such as remote homology detection and functional assignments. As a second example, we also discuss the problem of predicting residue–residue contacts in proteins. In this case, the task is much more complex than secondary structure prediction, and no satisfactory results have been achieved so far. Differently from the secondary structure sub-problem, the residue–residue contact sub-problem is not intrinsically simpler than the prediction of the protein structure, since a roughly correctly predicted set of residue–residue contacts would directly lead to prediction of a protein backbone very close to the real structure. These two protein structure sub-problems are discussed in the light of the current evaluation of the performance that are based on periodical blind-checks (CASP meetings) and permanent evaluation (EVA servers).