Assessment and refinement of eukaryotic gene structure prediction with gene-structure-aware multiple protein sequence alignment

Abstract

BackgroundAccurate computational identification of eukaryotic gene organization is a long-standing problem. Despite the fundamental importance of precise annotation of genes encoded in newly sequenced genomes, the accuracy of predicted gene structures has not been critically evaluated, mostly due to the scarcity of proper assessment methods.ResultsWe present a gene-structure-aware multiple sequence alignment method for gene prediction using amino acid sequences translated from homologous genes from many genomes. The approach provides rich information concerning the reliability of each predicted gene structure. We have also devised an iterative method that attempts to improve the structures of suspiciously predicted genes based on a spliced alignment algorithm using consensus sequences or reliable homologs as templates. Application of our methods to cytochrome P450 and ribosomal proteins from 47 plant genomes indicated that 50 ~ 60 % of the annotated gene structures are likely to contain some defects. Whereas more than half of the defect-containing genes may be intrinsically broken, i.e. they are pseudogenes or gene fragments, located in unfinished sequencing areas, or corresponding to non-productive isoforms, the defects found in a majority of the remaining gene candidates can be remedied by our iterative refinement method.ConclusionsRefinement of eukaryotic gene structures mediated by gene-structure-aware multiple protein sequence alignment is a useful strategy to dramatically improve the overall prediction quality of a set of homologous genes. Our method will be applicable to various families of protein-coding genes if their domain structures are evolutionarily stable. It is also feasible to apply our method to gene families from all kingdoms of life, not just plants.

Expanding the view on the evolution of the nematode dauer signalling pathways: refinement through gene gain and pathway co-option

“…Importantly, the sole reliance on gene models would have resulted in the homologous relationships for over 30 % of the genes going unfound. Central to this is the new crop of protein-to-genome aligners that are refining long established genome annotations [ 33 , 66 ].…”

MAFFT online service: multiple sequence alignment, interactive sequence choice and visualization

“…Recently, we have access to huge amounts of sequence data from widely divergent organisms, but the quality of the data is not always high because of the limitations of sequencing technologies. In the case of amino acid sequence data, the difficulty in eukaryotic gene prediction [28–30] also results in errors in data. It might be possible to automatically exclude such problematic data in certain cases, but sometimes, biologically important information is in low-quality sequences, especially when interest is in nonmodel organisms.…”

A simple method to control over-alignment in the MAFFT multiple sequence alignment program

“…Unfortunately, the quality of the sequences is not always high, partly due to limitations in sequencing technologies. Moreover, at the amino acid sequence level, a number of errors can be introduced due to difficulty in gene prediction ( Brent, 2005 ; Gotoh et al , 2014 ; Nagy and Patthy, 2013 ; Yandell and Ence, 2012 ). With incorrect reading frames, unrelated amino acid segments can appear in a set of homologous sequences.…”