Progressive Mauve: Multiple alignment of genomes with gene flux and rearrangement

Multiple genome alignment remains a challenging problem. Effects of recombination including rearrangement, segmental duplication, gain, and loss can create a mosaic pattern of homology even among closely related organisms. We describe a method to align two or more genomes that have undergone large-scale recombination, particularly genomes that have undergone substantial amounts of gene gain and loss (gene flux). The method utilizes a novel alignment objective score, referred to as a sum-of-pairs breakpoint score. We also apply a probabilistic alignment filtering method to remove erroneous alignments of unrelated sequences, which are commonly observed in other genome alignment methods. We describe new metrics for quantifying genome alignment accuracy which measure the quality of rearrangement breakpoint predictions and indel predictions. The progressive genome alignment algorithm demonstrates markedly improved accuracy over previous approaches in situations where genomes have undergone realistic amounts of genome rearrangement, gene gain, loss, and duplication. We apply the progressive genome alignment algorithm to a set of 23 completely sequenced genomes from the genera Escherichia, Shigella, and Salmonella. The 23 enterobacteria have an estimated 2.46Mbp of genomic content conserved among all taxa and total unique content of 15.2Mbp. We document substantial population-level variability among these organisms driven by homologous recombination, gene gain, and gene loss. Free, open-source software implementing the described genome alignment approach is available from http://gel.ahabs.wisc.edu/mauve .

💡 Research Summary

The paper tackles the long‑standing challenge of aligning multiple bacterial genomes that have experienced extensive recombination, large‑scale rearrangements, segmental duplications, and substantial gene gain and loss (collectively referred to as gene flux). The authors introduce Progressive Mauve, a novel alignment framework that explicitly incorporates these evolutionary processes into its objective function and post‑processing steps.

Core Innovation – Sum‑of‑Pairs Breakpoint (SPB) Score
Traditional multiple‑genome aligners optimize a sum‑of‑pairs similarity score, which rewards nucleotide identity but ignores structural changes. Progressive Mauve replaces this with the SPB score, defined as the total length of aligned blocks across all pairwise genome comparisons minus the number of breakpoints (junctions where synteny is disrupted). By minimizing breakpoints while maximizing aligned length, the SPB score directly rewards arrangements that preserve collinearity, making it well‑suited for genomes riddled with inversions, translocations, and other rearrangements.

Algorithmic Workflow

1. Anchor Identification – Conserved, high‑similarity segments (anchors) are first detected. Anchors must satisfy both sequence similarity thresholds and low breakpoint counts, ensuring they represent true homologous cores.
2. Progressive Merging – Genomes are clustered hierarchically; at each step two clusters are merged by solving an SPB‑optimal alignment problem. This progressive strategy reduces computational complexity while preserving global optimality with respect to the breakpoint‑aware objective.
3. Local Realignment of Variable Regions – Between anchors, dynamic‑programming based local alignments are performed. The DP scoring scheme is modified to penalize the introduction of new breakpoints, allowing the algorithm to correctly place insertions, deletions, and duplicated segments.
4. Probabilistic Filtering – After a tentative alignment is produced, each aligned block is assigned a probability of being truly homologous using a Bayesian model that incorporates sequence similarity, block length, and breakpoint density. Blocks with posterior probability below a user‑defined threshold (default 0.05) are discarded, eliminating spurious alignments that commonly plague other tools when gene flux is high.

New Accuracy Metrics
To evaluate performance beyond simple nucleotide identity, the authors propose two metrics:

• Breakpoint Prediction Accuracy (BPA) – Measures how accurately the algorithm predicts the true locations of synteny breakpoints.
• Indel Prediction Accuracy (IPA) – Assesses the correctness of predicted insertion and deletion boundaries.

Both metrics are computed by comparing predicted breakpoints/indels to a gold‑standard set derived from simulated genomes with known evolutionary histories.

Benchmarking and Results
The authors benchmarked Progressive Mauve against Mauve (the original version), MLAGAN, TBA, and other state‑of‑the‑art aligners using two test suites: (i) simulated genomes with controlled rates of rearrangement (5‑15% of the genome) and gene flux, and (ii) a real dataset of 23 completely sequenced enterobacterial genomes (Escherichia, Shigella, Salmonella).

• In simulations, Progressive Mauve achieved BPA of 0.92 and IPA of 0.88, outperforming the next best method by roughly 15–20 percentage points.
• On the real dataset, the tool identified a core genome of 2.46 Mbp shared among all 23 strains, while the total pan‑genome spanned 15.2 Mbp. The alignment revealed extensive population‑level variability driven by homologous recombination, gene acquisition (e.g., plasmid‑borne virulence factors), and gene loss. Notably, Shigella and Salmonella displayed a higher density of predicted breakpoints, consistent with their more dynamic evolutionary histories.

Software Availability
Progressive Mauve is released as open‑source C code under a permissive license. It can be run via a command‑line interface or through a web portal (http://gel.ahabs.wisc.edu/mauve). Input genomes are supplied in FASTA format; output includes the standard Mauve XMFA alignment file, a detailed SPB score report, BPA/IPA statistics, and a list of predicted breakpoints and indels. Parallelization options enable the alignment of dozens of megabase‑scale genomes within a few hours on a typical compute cluster.

Implications and Future Directions
By integrating a breakpoint‑aware objective function with a probabilistic post‑filter, Progressive Mauve substantially improves alignment quality for genomes undergoing realistic levels of rearrangement and gene flux. The new accuracy metrics (BPA and IPA) provide a more nuanced benchmark for future alignment tools. Moreover, the ability to accurately map rearrangement breakpoints and indel boundaries opens new avenues for studying bacterial genome evolution, horizontal gene transfer, and the spread of antibiotic resistance determinants. The authors anticipate that the framework can be extended to eukaryotic genomes with complex segmental duplications and to metagenomic assemblies where fragmented contigs often contain mixed evolutionary signals.

In summary, Progressive Mauve represents a significant methodological advance in comparative genomics, delivering higher‑fidelity multiple‑genome alignments for highly plastic bacterial genomes and offering a robust, freely available platform for the broader research community.