The SIEVE Server is a public web tool for prediction of type III secreted effectors, a refined version of the tool first described here (Samudrala et al. PLoS Pathogens, 2009) and more recently reviewed here (McDermott et al. Infection & Immunity, 2011). The SIEVE Server scores potential secreted effectors from genomes of bacterial pathogens with type III secretion systems using a model learned from known secreted proteins. The SIEVE Server requires only protein sequences of proteins to be screened and returns a conservative probability that each input protein is a type III secreted effector.
Many bacterial pathogens utilize a mechanism known as type III secretion systems to translocate proteins from the pathogen to the host. These tranlocated proteins (i.e. effectors) then disrupt and/or modify the host cell's response to a bacterial threat making the environment more conducive to infection. The SIEVE algorithm has been trained and tested on several bacterial pathogens, however it should be noted that the signal for type III secretion systems is highly cryptic with very little if any sequence identity among effectors, hence the need for a computational method to score possible secretion. Accordingly, the output from SIEVE should be treated as a prediction of secretion, but experimental validation is needed to fully demonstrate secretion.
Salmonella and Yersinia SIEVE
Our algorithm requires a large sequence similarity search using BLAST. To accommodate the efficient processing of large sets of protein sequences, for example whole genome sequences, we use ScalaBLAST, a parallel implementation of BLAST (Oehmen and Nieplocha. IEEE Trans on Para & Dist Sys, 2006), currently running on a 68 processor cluster. ScalaBLAST is used to generate the phylogenetic profile necessary for input into our computational model. For example, using this framework for a submitted fasta file with 7230 protein sequences (from the genome of Burkholderia pseudomallei) finishes in 28 minutes.
The SIEVE Server web tool is freely available with a brief user registration at this link "Biopilot Login for Access to SIEVE Server".