Most proteins comprise one or several domains. New domain architectures can be created by combining previously existing domains. The elementary events that create new domain architectures may be categorized into three classes, namely domain(s) insertion or deletion (indel), exchange and repetition. Using 'DomainTeam', a tool dedicated to the search for microsyntenies of domains, we quantified the relative contribution of these events. This tool allowed us to collect homologous bacterial genes encoding proteins that have obviously evolved by modular assembly of domains. We show that indels are the most frequent elementary events and that they occur in most cases at either the N- or C-terminus of the proteins. As revealed by the genomic neighbourhood/context of the corresponding genes, we show that a substantial number of these terminal indels are the consequence of gene fusions/fissions. We provide evidence showing that the contribution of gene fusion/fission to the evolution of multi-domain bacterial proteins is lower-bounded by 27% and upper-bounded by 64%. We conclude that gene fusion/fission is a major contributor to the evolution of multi-domain bacterial proteins.

Fast sequence databanks search algorithms generally make use of hash tables and look for exactly matching words. An increased sensitivity--at the expense of a decreased selectivity--can be attained in the case of proteins by using a reduced amino acid alphabet. We propose here an alphabet reduced to 10 symbols, that we used in modified versions of the FASTP and SCAN programs. An application to the aminoacyl-tRNA synthetases shows that this technique may be useful in detecting distant relationships between proteins.

A method aimed at classifying protein sequences without resorting to pairwise alignment is presented. Called DOCMA (DOt-plot Comparisons by Multivariate Analysis), it is based on a multivariate analysis of the pairwise dot-plots between all the sequences in the set. The dot-plots are first simplified by considering only the projections of the 'diagonal' segments of similarity onto the axes. From these projections a data matrix is built, in which each column is representative of the comparisons of one given sequence with all the other ones. This data matrix is then transformed into a distance matrix by a chi-squared analysis, from which the coordinates of the sequences in an orthonormal Euclidean space are obtained. The sequences are finally classified by a dynamic clustering procedure followed by a search for strong clusters. Application of this method to protein families such as the globins, the cytochromes c and the aminoacyl-tRNA synthetases shows that it is quite effective in delineating subgroups that contain even distantly related sequences.