Evaluating the usefulness of alignment filtering methods to reduce the impact of errors on evolutionary inferences

Arnaud Di Franco,Raphaël Poujol,Denis Baurain,Hervé Philippe

doi:10.1186/s12862-019-1350-2

Abstract

BackgroundMultiple Sequence Alignments (MSAs) are the starting point of molecular evolutionary analyses. Errors in MSAs generate a non-historical signal that can lead to incorrect inferences. Therefore, numerous efforts have been made to reduce the impact of alignment errors, by improving alignment algorithms and by developing methods to filter out poorly aligned regions. However, MSAs do not only contain alignment errors, but also primary sequence errors. Such errors may originate from sequencing errors, from assembly errors, or from erroneous structural annotations (such as incorrect intron/exon boundaries). Even though their existence is acknowledged, the impact of primary sequence errors on evolutionary inference is poorly characterized.ResultsIn a first step to fill this gap, we have developed a program called HmmCleaner, which detects and eliminates these errors from MSAs. It uses profile hidden Markov models (pHMM) to identify sequence segments that poorly fit their MSA and selectively removes them. We assessed its performances using > 700 amino-acid MSAs from prokaryotes and eukaryotes, in which we introduced several types of simulated primary sequence errors. The sensitivity of HmmCleaner towards simulated primary sequence errors was > 95%. In a second step, we compared the impact of segment filtering software (HmmCleaner and PREQUAL) relative to commonly used block-filtering software (BMGE and TrimAI) on evolutionary analyses. Using real data from vertebrates, we observed that segment-filtering methods improve the quality of evolutionary inference more than the currently used block-filtering methods. The formers were especially effective at improving branch length inferences, and at reducing false positive rate during detection of positive selection.ConclusionsSegment filtering methods such as HmmCleaner accurately detect simulated primary sequence errors. Our results suggest that these errors are more detrimental than alignment errors. However, they also show that stochastic (sampling) error is predominant in single-gene evolutionary inferences. Therefore, we argue that MSA filtering should focus on segment instead of block removal and that more studies are required to find the optimal balance between accuracy improvement and stochastic error increase brought by data removal.

Highlights

Multiple Sequence Alignments (MSAs) are the starting point of molecular evolutionary analyses
A profile hidden Markov models (pHMM) is built from the MSA using HMMER (Fig. 1a); it will be used as the reference, i.e., the underlying model having generated each sequence of the MSA
Each sequence of the MSA is evaluated with the pHMM (Fig. 1b), which yields one profile-sequence alignment per sequence through the heuristic of HMMER [19]

Summary

Introduction

Multiple Sequence Alignments (MSAs) are the starting point of molecular evolutionary analyses. MSAs do contain alignment errors, and primary sequence errors Such errors may originate from sequencing errors, from assembly errors, or from erroneous structural annotations (such as incorrect intron/exon boundaries). Even though their existence is acknowledged, the impact of primary sequence errors on evolutionary inference is poorly characterized. Di Franco et al BMC Evolutionary Biology (2019) 19:21 they generate a non-phylogenetic signal, conflicting with the genuine (historical) phylogenetic signal in the data [2, 3]. Their presence inflates estimates of positive selection [4, 5]. Some studies suggest that block-filtering software improves evolutionary inference [13,14,15,16], whereas other authors find support for the opposite [17, 18]

Methods

Results

Conclusion