An important mechanism for the evolution of toxins in venomous animals is believed to be the acquisition of genes encoding proteins that switch from physiological to toxic roles following gene duplication. The 'reverse recruitment' hypothesis pertains that these genes can also revert back to physiological functions, although such events are thought to be rare. A non-supervised homology searching method was developed which allowed the peptide diversity of animal toxins to be described as combinations between limited numbers of amino-acid sequence blocks we called 'tox-bits'. Taking the phospholipase A2 (PLA2) protein family as an example, a Bernoulli Trial was used to test if 'tox-bits' were robust enough to distinguish between peptides with physiological or toxin functions. The analysis revealed that discrimination was indeed possible, and supports the very recent 'restriction' hypothesis whereby genes with the potential to encode toxic functions have likely been independently recruited into venom systems and therefore require few, if any, reverse recruitment events. The development of 'tox-bits' provides a novel bioinformatics tool to allow recognition of toxins from other proteins in genome sequences, facilitating the study of gene recruitment and duplication strategies in venom diversification. The 'tox-bits' library is freely available at http://bioserv.pbf.hr/blocks.zip.
Keywords: Evolution; Hidden Markov models; Multiple alignments; Toxin diversification.
Copyright © 2015 Elsevier Ltd. All rights reserved.