FastTree

What is FastTree?

FastTree 2.2 infers approximately maximum-likelihood phylogenetic trees from aligned protein or nucleotide sequences. It is designed for analyses where conventional maximum-likelihood tree searches would be too slow, while still refining the topology and branch lengths with likelihood-based methods.

FastTree requires a multiple sequence alignment. Every sequence must contain the same number of aligned positions. If the sequences are not aligned yet, first run Clustal Omega or MUSCLE5.

How FastTree works

FastTree builds and refines a tree in three stages:

1. It constructs an initial topology with a profile-based, heuristic form of neighbor joining.
2. It improves that topology with minimum-evolution nearest-neighbor interchange (NNI) and subtree-pruning-regrafting (SPR) moves.
3. It performs maximum-likelihood NNI refinement and optimizes branch lengths under the selected substitution model.

The default CAT approximation assigns sites to rate categories during tree inference. Enabling Gamma optimization recalculates the final likelihood and branch lengths with a Gamma20 model, but does not redo the topology search under gamma rates.

Input requirements

Submit one aligned FASTA or PHYLIP file, up to 20 MB. Set PHYLIP alignment count above 1 to analyze a concatenated stream of PHYLIP alignments. Every native tree or distance matrix is retained. FASTA inputs may use - or . for gaps. Alignments with one or two sequences are accepted because FastTree itself supports them, although meaningful phylogenetic inference usually requires more sequences.

FastTree treats the input as protein by default. Select Nucleotide explicitly for DNA or RNA alignments; sequence type is not inferred automatically.

Sequence names

Without Quote sequence names, FastTree ends a sequence name at whitespace or at (, ), ,, or :. Names that become identical after that truncation are rejected before the job starts.

Enable Quote sequence names to retain those characters and quote names in the Newick tree. Native FastTree ends quoted FASTA names at an apostrophe or tab. PHYLIP follows its own name parser. Blank lines and spaces before the first > are removed before FastTree runs, because native FastTree refuses FASTA that does not start with >.

Evolutionary models

Protein models

• JTT is the native default.
• WAG and LG provide alternative empirical amino-acid models.

Nucleotide models

• Jukes-Cantor (JC) is the native nucleotide default.
• GTR estimates distinct substitution rates and base frequencies. Enable it when that added model complexity is appropriate for the alignment.

Settings

Gamma optimization

Uses Gamma20 likelihoods to optimize final branch lengths and report a gamma-corrected likelihood. This is slower than the default CAT-only result.

Fast mode

Disables local hill climbing during the initial join search. It can substantially reduce runtime for very large alignments, with a possible loss of tree-search accuracy.

Exhaustive join search

Compares every candidate join during initial tree construction. It is more thorough and substantially slower.

Distance pseudocounts

Pseudocounts stabilize distance estimates for sequences with little or no overlapping, non-gap data. The default native weight is 1; higher values increase the contribution of the prior distance estimate.

Local support values

FastTree reports SH-like local support values by default. Support resamples controls the number of site-likelihood resamples used to calculate those values; the native default is 1,000. This is not traditional bootstrap analysis because FastTree does not rebuild a complete tree for every resample.

Disable Compute local support values to omit internal-node support calculations. Zero support resamples also disables support. The support seed accepts zero and negative integers and makes repeated analyses with the same inputs and settings reproducible.

Native inputs and advanced settings

Input	Format	When required
Starting tree	Newick text	Use starting tree
Topology constraints	FASTA or PHYLIP with native 0, 1, and - states	Use topology constraints
Custom distance model	Three native files: .distances, .inverses, .eigenvalues	Custom protein distance model
Protein transition matrix	Native FastTree matrix text	Custom protein transition matrix

Reuse first starting tree applies the first supplied topology to every alignment. Otherwise FastTree reads one starting tree per alignment. A saved .trees collection can be supplied directly as the starting-tree file. Constraint weight defaults to 100 and must be positive. Native file contents are passed unchanged.

Parameter	Type	Default	Description
Analysis mode	enum	tree	Infer trees or calculate pairwise distance matrices.
PHYLIP alignment count	integer	1	Number of alignments to read from one PHYLIP stream.
Distance correction	enum	log	Log-corrected or raw native distances.
Protein distance model	enum	blosum45	BLOSUM45, percent different, or a supplied custom model.
Fixed GTR rates	boolean	false	Supply AC, AG, AT, CG, CT, GT rates; each defaults to 1 and must be at least 0.00001.
Fixed GTR frequencies	boolean	false	Supply A, C, G, T frequencies; each defaults to 0.25 and must be at least 0.00001. Their sum must be within 0.01 of 1; FastTree normalizes them.
CAT rate categories	integer	20	Use 1 to disable rate variation across sites.
BIONJ weighted joins	boolean	false	Use weighted joins during native initial tree construction.
Disable minimum-evolution refinement	boolean	false	Skip ME NNI and SPR refinement.
Minimum-evolution NNI rounds	integer	-1	Native automatic budget; 0 disables NNI and default SPR rounds.
SPR rounds / maximum SPR length	integer	2 / 10	Native refinement budgets.
Maximum-likelihood refinement	enum	search	Topology search, branch-length-only optimization, or minimum-evolution-only inference.
ML NNI rounds	integer	-1	Native automatic budget, applicable to ML topology search.
ML length optimization rounds	integer	1	Native -mlacc refinement control.
Exhaustive NNI moves	boolean	false	Consider all native NNI moves.
Approximate ML posteriors	boolean	false	Use native approximate posteriors; exact is the source default.
Top hits	enum	default	Native default, enabled, disabled, or a custom multiplier.
Top hits multiplier	number	1	Used with custom top hits.
Close-neighbor fraction	number	0.75, inactive	Optional override, strictly between 0 and 1; requires positive top hits.
Top-hits refresh fraction	number	0.8, inactive	Optional override, strictly between 0 and 1; requires positive top hits.
Native diagnostic verbosity	integer	1	Native diagnostic detail; 0 is quiet. Higher levels retain additional reports.
Suppress progress status	boolean	false	Omit native progress updates while retaining other diagnostics.
Second-level top hits	enum	default	Native default, explicitly enabled, or explicitly disabled.

Integer count controls use the native signed 32-bit range. Seeds use the exact JavaScript integer range. Pseudocount weights retain their submitted precision and have no artificial upper cap. The hosted upload, memory, and time limits still apply.

Results

A single tree run returns a downloadable Newick tree rendered in the interactive tree viewer. Multiple-alignment runs return the complete native Newick tree collection. Distance-matrix mode returns all native matrices in PHYLIP distance-matrix format. Those collections and matrices are available in Files. Branch lengths represent expected substitutions per site, and internal-node labels contain SH-like support values when support calculation is enabled.

Successful runs also include the native optimization log and diagnostics. Failed or timed-out native runs retain complete available diagnostics, partial stdout, and optimization logs in Files.

The result also includes:

• sequence, unique-sequence, and alignment-position counts;
• the effective model and sequence type;
• final log-likelihood and total branch length when reported by FastTree;
• the native optimization log with intermediate trees and model details;
• the complete native progress and warning report.

Runtime and limitations

Hosted jobs use an 8 GB compute environment and may run for up to about 58 minutes. Large or diverse alignments can still exceed that envelope; reduce the alignment or enable Fast mode if a job times out. The 20 MB input limit is a service limit, not a limit of the FastTree source program.

FastTree prioritizes speed over an exhaustive maximum-likelihood search. For final analyses where small topology differences are critical, compare the result with a more thorough method such as IQ-TREE or RAxML-NG.