Classifying Sequences by the Optimized Dissimilarity Space Embedding Approach: a Case Study on the Solubility Analysis of the E. coli Proteome.