MUSCLE5

What is MUSCLE5?

MUSCLE5 is a multiple sequence alignment method for proteins and nucleic acids. Version 5 introduced improved guide-tree construction, the high-accuracy PPP algorithm, a scalable Super5 mode, and alignment ensembles for measuring uncertainty.

ProteinIQ runs the verified MUSCLE 5.3 Linux release. MUSCLE detects the sequence alphabet itself, matching the source program's default behavior.

Choose an alignment mode

MUSCLE does not automatically switch from PPP to Super5. Select Super5 explicitly when runtime or dataset size makes PPP impractical.

Single alignments and ensembles

The default result is one alignment. PPP mode also supports two native ensemble strategies:

• Stratified ensemble runs the requested number of perturbations for each of four guide-tree permutations. The MUSCLE default is 4 replicates per permutation, or 16 alignments in total.
• Diversified ensemble combines guide-tree permutations and perturbations to produce a larger set of alternatives. The MUSCLE default is 100 alignments.

Ensemble members have approximately comparable expected accuracy; their order is not a quality ranking. ProteinIQ also runs MUSCLE's native ensemble-analysis commands so you can inspect dispersion, select the highest-total-column-confidence member, and map column or letter confidence without discarding any member.

Ensembles consume compute in proportion to the number of alignments generated, so the credit estimate scales by the requested ensemble size.

Input requirements

Submit one FASTA record or a multi-record FASTA file. Each record begins with > and is followed by one or more sequence lines. MUSCLE accepts an empty label, although descriptive unique labels are strongly recommended for interpreting results.

>human_insulin
MALWMRLLPLLAVTFLAGCGAKSQVQLVESGGGLVQPGGSLRLSCAASGFTFSGYY
>mouse_insulin
MALWMRLLPLLAVTFLAGCGAKSSVQLLESGGGLVQPGGSLRLSCAASGFTFSGYY
>zebrafish_insulin
MQLWMRLPPLAVTFLVLCGAKSSVQLVESGGGLVQPGGSLRLSCAASGFTFSGYY

MUSCLE 5.3 accepts a single sequence and returns it unchanged as a valid one-record alignment. Protein, DNA, and RNA inputs are supported. Do not mix different alphabets in one job.

Settings

Results and files

A single-alignment job returns:

• an aligned FASTA (.afa) file used by the MSA viewer and available to downstream workflows;
• a native diagnostic log containing the MUSCLE version, progress, and warnings.

When All four is selected in Super5, all four aligned FASTA variants are returned, billed, and available to downstream workflows. The native none.0 variant opens first in the viewer; the variants are alternatives, not a quality ranking.

An ensemble job additionally returns:

• every member as an individual aligned FASTA file;
• the combined native ensemble FASTA (.efa) archive;
• a table containing MUSCLE's D_LP, D_Cols, CC minimum, average, maximum, and best-member label;
• the highest-total-column-confidence alignment (maxcc.afa);
• a confidence-sequence-annotated EFA archive (confidence.efa);
• a letter-confidence AFA and matching Jalview feature file;
• the complete diagnostic log for alignment, extraction, dispersion, and confidence analysis.

The first ensemble member opens in the MSA viewer for convenience. It is not a best-scoring or recommended member; use the separate MaxCC alignment when you specifically need MUSCLE's highest-total-column-confidence member. In the confidence EFA, two digits encode column confidence (for example, 73 means 0.73 and ++ means 1.0). The letter-confidence AFA uses digits 0-9 in place of residues and is intended as an annotation companion to the MaxCC alignment.

Runtime and limits

Jobs run with 8 CPU cores and 8 GB of memory. The native process can run for up to 115 minutes within a 120-minute job limit. Diversified ensembles and large PPP alignments can take substantially longer than one alignment; use Super5 or reduce the dataset if a job approaches the limit.

Accepted file extensions are .fasta, .fa, .fas, and .txt, with a 50 MB upload limit. Account-level text and job limits also apply.

To keep multi-alignment result bundles reliable, the estimated and generated combined EFA archive is limited to 50 MB. The complete returned bundle—including split alignments, confidence artifacts, Jalview features, and diagnostics—is limited to 100 MB and conservatively estimated before compute. Reduce the input size or alignment count if a request would exceed either output budget.

Use cases

• Build an alignment before phylogenetic inference.
• Locate conserved residues and motifs across homologs.
• Prepare aligned sequences for structure or function analysis.
• Test whether downstream conclusions are stable across alternative alignments.

For a different speed-accuracy tradeoff or additional output formats, use MAFFT or Clustal Omega. For database-scale sequence search, use MMseqs2.

Limitations

• Progressive alignment errors made early in guide-tree traversal can affect later columns.
• Very divergent sequence families may have multiple equally plausible alignments.
• This sequence tool does not expose MUSCLE's separate MEGA/Reseek structure-alignment workflow or infer structures and functional annotations.
• PPP ensembles are uncertainty samples, not ranked candidate alignments.
• Ensemble resampling, standalone EFA conversion, the broken v5.3 HTML confidence branch, and low-level consistency/refinement controls are not available in the hosted form.

Reference

Edgar RC. MUSCLE v5 enables improved estimates of phylogenetic tree confidence by ensemble bootstrapping. Nature Communications 13, 6968 (2022). https://doi.org/10.1038/s41467-022-34630-w