FragGeneScan: predicting genes in short and error-prone reads

Abstract

The advances of next-generation sequencing technology have facilitated metagenomics research that attempts to determine directly the whole collection of genetic material within an environmental sample (i.e. the metagenome). Identification of genes directly from short reads has become an important yet challenging problem in annotating metagenomes, since the assembly of metagenomes is often not available. Gene predictors developed for whole genomes (e.g. Glimmer) and recently developed for metagenomic sequences (e.g. MetaGene) show a significant decrease in performance as the sequencing error rates increase, or as reads get shorter. We have developed a novel gene prediction method FragGeneScan, which combines sequencing error models and codon usages in a hidden Markov model to improve the prediction of protein-coding region in short reads. The performance of FragGeneScan was comparable to Glimmer and MetaGene for complete genomes. But for short reads, FragGeneScan consistently outperformed MetaGene (accuracy improved ∼62% for reads of 400 bases with 1% sequencing errors, and ∼18% for short reads of 100 bases that are error free). When applied to metagenomes, FragGeneScan recovered substantially more genes than MetaGene predicted (>90% of the genes identified by homology search), and many novel genes with no homologs in current protein sequence database.

Comparative study of viromes from freshwater samples of the Ile-Balkhash region of Kazakhstan captured through metagenomic analysis

“…Before uploading, sequences were quality trimmed using MG-RAST QC pipeline, which included removal of artificial or technical replicates [16] and removal of low quality sequences [17]. Gene-calling was performed using MG-RAST automated pipeline which included the use of FragGeneScan [18], clustering of predicted proteins at 90% identity by using uclust [19] and the use of sBLAT, and applying of the BLAT algorithm [20] for similarity analysis of each cluster. Taxonomic composition of the viromes was obtained through comparison of sequences to the curated NCBI RefSeq complete viral genomes protein sequence database using blastX with an e-value cut-off of B10 -3 .…”

Ecological succession leads to chemosynthesis in mats colonizing wood in sea water

“…FragGeneScan (Rho et al, 2010) identified openreading frames, which were annotated by BLASTX (Camacho et al, 2009) against the KEGG (Kanehisa, 2002) and SILVA (Pruesse et al, 2007) databases. Analyses were conducted with the MG-RAST pipeline (Meyer et al, 2008) and sequences were deposited under the accession numbers 4623131.3, 4623132.3 and 4623133.3.…”

Automated and accurate estimation of gene family abundance from shotgun metagenomes

“…While 6FT captures all true open reading frames (ORFs), it also increases the volume of query sequences and can lead to classification of spurious ORFs. Alternatively, ab initio metagenomic gene prediction [26], [27], [28] decreases the volume of queries, but can result in false-positive and false-negative reading frame predictions Metagenomic simulation framework. 1) Taxonomic profiles from real metagenomes are used to construct mock microbial communities.…”

“…ORFs were searched and classified into SFams using RAPsearch2 while optimizing all other parameters (Methods). We ran 6FT in addition to three different metagenomic gene predictors-FragGeneScan [27], MetaGeneMark [28], and Prodigal [26]-and classified reads into protein families according to their top-hit across predicted ORFs (per-read annotation). Additionally, we evaluated a novel heuristic to rapidly filter short spurious ORFs (S2 Fig). We found that the metagenomic gene finders reduced sequence volume by~85% relative to ORFs that had been naively translated in 6-frames (Fig 2A), consistent with prior observations [29].…”

“…In the first, we performed naïve translation of reads into all 6 possible reading frames (6FT) using the Transeq application from EMBOSS (v6.3.1) [82]. In the second approach, we took ORFs predicted from 6FT, split the sequences on stop-codons, and subsequently filtered these sequences to remove short spurious ORFs and reduce data volume ( S2 Fig). In the third approach, we used three popular tools-FragGeneScan [27], MetaGeneMark [28], and Prodigal [26]-to perform ab initio gene prediction. FragGeneScan (v1.18) was run using the options: -complete = 0 -train = illumina_10; MetaGeneMark (v3.25) was run using the options: -m MetaGeneMark_v1.mod; Prodigal (v2.60) was run using the options: -p meta.…”