Machine learning and the prediction of protein structure: the state of the art

In the genomic era machine learning algorithms that improve automatically through experience have proven to be among the most successful methods for addressing relevant problems of Computational Molecular Biology, including protein structure prediction. The increasing amount of information stored in publicly available biological data bases is retrieved to find approximate solutions relating sequence to protein structure. This may be useful in different fields of Bioinformatics, from structural, functional and comparative genomics, to protein engineering and molecular medicine. How far can we go if we have a protein sequence and we do not know the corresponding structure? Also, why is it so important to know the protein structure? This and related issues will be discussed in the following