A poor man's BLASTX--high-throughput metagenomic protein database search using PAUDA

Abstract

Summary: In the context of metagenomics, we introduce a new approach to protein database search called PAUDA, which runs ~10,000 times faster than BLASTX, while achieving about one-third of the assignment rate of reads to KEGG orthology groups, and producing gene and taxon abundance profiles that are highly correlated to those obtained with BLASTX. PAUDA requires <80 CPU hours to analyze a dataset of 246 million Illumina DNA reads from permafrost soil for which a previous BLASTX analysis (on a subset of 176 million reads) reportedly required 800,000 CPU hours, leading to the same clustering of samples by functional profiles.

Availability: PAUDA is freely available from: http://ab.inf.uni-tuebingen.de/software/pauda. Also supplementary method details are available from this website.

Fig. 1.

An overview of the PAUDA approach

Fig. 2.

KEGG comparison of PAUDA and BLASTX. Left: Each true KO group is represented by a dot with coordinates that correspond to the number of reads assigned to the KO group by BLASTX (on the x-axis) and PAUDA (on the y-axis). Right: To show the low abundance KO groups more clearly, here, we plot the same data on a logarithmic scale