ReformAlign: improved multiple sequence alignments using a profile-based meta-alignment approach

Abstract

BackgroundObtaining an accurate sequence alignment is fundamental for consistently analyzing biological data. Although this problem may be efficiently solved when only two sequences are considered, the exact inference of the optimal alignment easily gets computationally intractable for the multiple sequence alignment case. To cope with the high computational expenses, approximate heuristic methods have been proposed that address the problem indirectly by progressively aligning the sequences in pairs according to their relatedness. These methods however are not flexible to change the alignment of an already aligned group of sequences in the view of new data, resulting thus in compromises on the quality of the deriving alignment. In this paper we present ReformAlign, a novel meta-alignment approach that may significantly improve on the quality of the deriving alignments from popular aligners. We call ReformAlign a meta-aligner as it requires an initial alignment, for which a variety of alignment programs can be used. The main idea behind ReformAlign is quite straightforward: at first, an existing alignment is used to construct a standard profile which summarizes the initial alignment and then all sequences are individually re-aligned against the formed profile. From each sequence-profile comparison, the alignment of each sequence against the profile is recorded and the final alignment is indirectly inferred by merging all the individual sub-alignments into a unified set. The employment of ReformAlign may often result in alignments which are significantly more accurate than the starting alignments.ResultsWe evaluated the effect of ReformAlign on the generated alignments from ten leading alignment methods using real data of variable size and sequence identity. The experimental results suggest that the proposed meta-aligner approach may often lead to statistically significant more accurate alignments. Furthermore, we show that ReformAlign results in more substantial improvement in cases where the starting alignment is of relatively inferior quality or when the input sequences are harder to align.ConclusionsThe proposed profile-based meta-alignment approach seems to be a promising and computationally efficient method that can be combined with practically all popular alignment methods and may lead to significant improvements in the generated alignments.Electronic supplementary materialThe online version of this article (doi:10.1186/1471-2105-15-265) contains supplementary material, which is available to authorized users.

The Quantification of Representative Sequences pipeline for amplicon sequencing: case study on within‐population ITS1 sequence variation in a microparasite infecting Daphnia

“…), Kalign (Lassmann & Sonnhammer ), MAFFT (Katoh & Standley ), MUSCLE (Edgar ), PicXAA (Sahraeian & Yoon ) and PRANK (Löytynoja & Goldman ). QRS can then (optionally) execute ReformAlign (Lyras & Metzler ) to postprocess the alignment based on a profile‐based meta‐alignment approach.…”

TPMA: A two pointers meta-alignment tool to ensemble different multiple nucleic acid sequence alignments

“…If the new alignment exhibits higher accuracy, it replaces the initial alignment. On the other hand, the horizontal realigners, including REFINER [ 5 ], RF [ 6 ], and ReformAlign [ 7 ], adopt distinct strategies. Similar to REFINER, RF randomly selects a sequence in each iteration and realigns it with the profile of the remaining sequences; if the accuracy of the new alignment improves, it serves as the input for the next iteration round until the objective function score converges or reaches the cycle limit.…”