Abstract
Motivation: The genomic era in molecular biology has brought on a rapidly widening gap between the amount of sequence data and first-hand experimental characterization of proteins. Fortunately, the theory of evolution provides a simple solution: functional and structural information can be transferred between homologous proteins. Sequence similarity searching followed by k-nearest neighbor classification is the most widely used tool to predict the function or structure of anonymous gene products that come out of genome sequencing projects.Results: We present a novel word filter, suffix array neighborhood search (SANS), to identify protein sequence similarities in the range of 50–100% identity with sensitivity comparable to BLAST and 10 times the speed of USEARCH. In contrast to these previous approaches, the complexity of the search is proportional only to the length of the query sequence and independent of database size, enabling fast searching and functional annotation into the future despite rapidly expanding databases.Availability and implementation: The software is freely available to non-commercial users from our website http://ekhidna.biocenter.helsinki.fi/downloads/sans.Contact: liisa.holm@helsinki.fi.
Highlights
The performance of suffix array neighborhood search (SANS) is close to the performance of USEARCH SANS is 10 times faster and USEARCH has the advantage of multiple testing
We have investigated the use of word filters to speed up protein sequence database searches
The principal conclusions from our extensive benchmarking can be summarized as follows: 1. word filters are as sensitive as BLAST in the feasible regime of 50–100% sequence identity, 2. many variants of word filters perform about well in the feasible regime but SANS is the most robust to parameter variation, 3. suffix array supports the fastest known word filter algorithm, 4. methods incorporating explicit alignment are necessary
Summary
2.1 Protein datasets We selected real datasets to get a realistic distribution of protein lengths, composition and protein family sizes (Table 2). Uniprot is the major collection of protein sequences. It consists of two parts, swissprot and trembl. Trembl contains protein sequences that are translated from nucleotide sequences and automatically annotated. The metagenome dataset is a collection of proteins detected in environmental samples and was downloaded from NCBI (env_nr). Metagenomic sequences typically come from uncultured organisms that are not present in the protein databases
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call