Orphelia: predicting genes in metagenomic sequencing reads

Abstract

Metagenomic sequencing projects yield numerous sequencing reads of a diverse range of uncultivated and mostly yet unknown microorganisms. In many cases, these sequencing reads cannot be assembled into longer contigs. Thus, gene prediction tools that were originally developed for whole-genome analysis are not suitable for processing metagenomes. Orphelia is a program for predicting genes in short DNA sequences that is available through a web server application (http://orphelia.gobics.de). Orphelia utilizes prediction models that were created with machine learning techniques on the basis of a wide range of annotated genomes. In contrast to other methods for metagenomic gene prediction, Orphelia has fragment length-specific prediction models for the two most popular sequencing techniques in metagenomics, chain termination sequencing and pyrosequencing. These models ensure highly specific gene predictions.

Environmental microbiology through the lens of high-throughput DNA sequencing: Synopsis of current platforms and bioinformatics approaches

“…The second method uses gene prediction algorithms for finding genes in shotgun HTS reads. These are implemented in software such as Orphelia, MetaGenemark, Fraggenescan and MGA (Hoff et al, 2009;Noguchi et al, 2008;Zhu et al, 2010) (Table 1). These tools all give relatively good predictions (>80%) on reads longer than 400 bp (Yok and Rosen, 2011).…”

FragGeneScan: Predicting Genes in Short and Error-Prone Reads

“…For the shorter reads Reads were simulated with 1% sequencing error rate for lengths of 100, 200 and 400 bp, and 0.5% sequencing error rate for length of 700 bp, respectively. The nine genomes are the same as those in Table 2, and were used for testing gene prediction in short reads (16,18). of 200 and 100 bps that have no sequencing errors, however, FragGeneScan outperforms MetaGene significantly.…”

“…In metagenomics, gene finding can provide the opportunity to elucidate the activities and interactions of genes within an environmental sample, from which the metabolic and signaling pathways specific to the environment can be reconstructed and identified (14). To date, only a few methods have been developed for gene prediction in metagenomic sequences (15)(16)(17)(18)(19). Most commonly, genes encoded by metagenomes have been identified by using homology-based methods such as BLASTX (20,21).…”

“…This way, novel families that have multiple members in a metagenomic dataset can be identified (22). The other straightforward solution to novel gene prediction in metagenomics is to use feature-based approaches such as probabilistic models to evaluate the probabilities of open reading frames (ORFs) being protein-coding regions (16,18,23), in a manner similar to conventional gene finding methods such as Glimmer and GeneMark (24)(25)(26).…”

On the Scalability of Supervised Learners in Metagenomics

“…There are many gene finding programs that are available for predicting genes in prokaryotic sequences. The Orphelia program [7] uses linear discriminants to decrease the number of features and then uses artificial neural networks to predict genes in the metagenomic sequence reads. In GeneMarkS [8], the authors claim that their method (hidden markov model) can be used for finding genes in prokaryotic genomes without prior knowledge of any proteins.…”