SUPER-FOCUS: a tool for agile functional analysis of shotgun metagenomic data

Abstract

Summary: Analyzing the functional profile of a microbial community from unannotated shotgun sequencing reads is one of the important goals in metagenomics. Functional profiling has valuable applications in biological research because it identifies the abundances of the functional genes of the organisms present in the original sample, answering the question what they can do. Currently, available tools do not scale well with increasing data volumes, which is important because both the number and lengths of the reads produced by sequencing platforms keep increasing. Here, we introduce SUPER-FOCUS, SUbsystems Profile by databasE Reduction using FOCUS, an agile homology-based approach using a reduced reference database to report the subsystems present in metagenomic datasets and profile their abundances. SUPER-FOCUS was tested with over 70 real metagenomes, the results showing that it accurately predicts the subsystems present in the profiled microbial communities, and is up to 1000 times faster than other tools.

Availability and implementation: SUPER-FOCUS was implemented in Python, and its source code and the tool website are freely available at https://edwards.sdsu.edu/SUPERFOCUS.

Contact: redwards@mail.sdsu.edu

Supplementary information: Supplementary data are available at Bioinformatics online.

Fig. 1.

Workflow of the SUPER-FOCUS program

Fig. 2.

Representation of a subsystem structure (Levels 1–3 classifications and Function)

Fig. 3.

Percent classification sensitivity (A) and precision (B) of level 1 subsystems and speed of RAPSearch2 and SUPER-FOCUS using different databases and parameter modes. This analysis was based on a comparison of 50 HMP metagenomes, where blastx assignments using DB_100 were considered to be the true answer

Fig. 4.

Percentage of level 3 subsystems present in all the testing set metagenomes predicted by SUPER-FOCUS

Fig. 5.

Confusion matrix displaying the percentage of correct assignments in each level 1 subsystem for the 50 HMP metagenomes. (a) Shows the RAPSearch2 assignments in the sensitive mode to DB_100. (b) Shows the SUPER-FOCUS assignments in the sensitive mode to DB_100

Fig. 6.

Classification sensitivity using level 1 classifications and speed comparison of 50 HMP metagenomes using RAPSearch2 and SUPER-FOCUS using different databases and modes, but removing Eurkaryota and viral assignments. blastx assignments using DB_100 were considered to be the true answer

Fig. 7.

Classification sensitivity (a) and precision (b) percent using level 1 and speed comparison of three viromes using RAPSearch2 and SUPER-FOCUS using different databases and modes. blastx assignments using DB_100 were considered to be the true answer

Fig. 8.

Run time comparison for the three marine viromes using SUPER-FOCUS, RTMg, MEGAN and MG-RAST

Fig. 9.

Run time comparison for the one big data metagenome using SUPER-FOCUS, RTMg, MEGAN and MG-RAST

Fig. 10.

Comparison of level 1 subsystems profile of one big data metagenome using SUPER-FOCUS, RTMg, MEGAN, MG-RAST and blastx that are considered to be the true answer

Fig. 11.

Box plots displaying the percent sensitivity (A and C) and precision (B and D) of RAPSearch2 (A and B), blastx (C and D) annotation of the 20 coral metagenomes. RAPsearch2 was tested in the fast and sensitive modes

Fig. 12.

Hierarchical clustering of the taxonomic (A) and functional (B) annotations of 20 coral metagenomes. Genus level taxonomic annotation was performed using FOCUS. Functional annotation of level 3 subsystems was performed using SUPER-FOCUS using blastx and DB_98