RAxML

RAxML (Randomized Axelerated Maximum Likelihood) is a tool for large-scale maximum likelihood phylogenetic analysis.

Basic Usage

# Standard analysis with automatic model selection
raxml-ng --all --msa alignment.fasta --model GTR+G --prefix output --threads 4

# With bootstrapping
raxml-ng --all --msa alignment.fasta --model GTR+G \
    --prefix output --threads 4 --bs-trees 100

Nucleotide Models

# Common DNA models
raxml-ng --msa alignment.fasta --model GTR+G      # General Time Reversible + Gamma
raxml-ng --msa alignment.fasta --model GTR+I+G    # + Invariant sites
raxml-ng --msa alignment.fasta --model HKY+G      # Hasegawa-Kishino-Yano
raxml-ng --msa alignment.fasta --model JC         # Jukes-Cantor
raxml-ng --msa alignment.fasta --model K80        # Kimura 2-parameter

# Automatic model selection
raxml-ng --msa alignment.fasta --model GTR+G --all \
    --tree pars{25},rand{25} --prefix output

Protein Models

# Protein analysis
raxml-ng --msa protein_alignment.fasta --model LG+G --prefix protein_tree

# Other protein models
raxml-ng --msa alignment.fasta --model WAG+G      # Whelan and Goldman
raxml-ng --msa alignment.fasta --model JTT+G      # Jones-Taylor-Thornton
raxml-ng --msa alignment.fasta --model DAYHOFF+G  # Dayhoff
raxml-ng --msa alignment.fasta --model LG+G+F     # + Empirical frequencies

Bootstrap Analysis

# Rapid bootstrap analysis
raxml-ng --bootstrap --msa alignment.fasta --model GTR+G \
    --prefix boot --threads 4 --bs-trees 1000

# Map bootstrap values to best tree
raxml-ng --support --tree best_tree.tree --bs-trees boot.raxml.bootstraps \
    --prefix final

# Automatic bootstrap convergence
raxml-ng --all --msa alignment.fasta --model GTR+G \
    --prefix output --bs-trees autoMRE --threads 4

Partitioned Analysis

# Create partition file (partitions.txt):
# GTR+G, part1 = 1-500
# GTR+G, part2 = 501-1000
# LG+G, part3 = 1001-1300

# Run with partitions
raxml-ng --msa alignment.fasta --model partitions.txt \
    --prefix partitioned --threads 4

Tree Search Strategies

# Multiple starting trees
raxml-ng --msa alignment.fasta --model GTR+G \
    --tree pars{10},rand{10} --prefix multi_start

# Starting from specific tree
raxml-ng --msa alignment.fasta --model GTR+G \
    --tree starting_tree.nwk --prefix from_tree

# Tree evaluation (no search)
raxml-ng --evaluate --msa alignment.fasta --model GTR+G \
    --tree input_tree.nwk --prefix eval

Python Interface

import subprocess
import os

def run_raxml(alignment_file, model='GTR+G', bootstraps=100,
              threads=4, prefix='raxml_output'):
    """Run RAxML-NG analysis."""

    cmd = [
        'raxml-ng', '--all',
        '--msa', alignment_file,
        '--model', model,
        '--prefix', prefix,
        '--threads', str(threads),
        '--bs-trees', str(bootstraps)
    ]

    result = subprocess.run(cmd, capture_output=True, text=True)

    if result.returncode == 0:
        # Parse results
        tree_file = f'{prefix}.raxml.bestTree'
        support_file = f'{prefix}.raxml.support'

        return {
            'best_tree': tree_file if os.path.exists(tree_file) else None,
            'support_tree': support_file if os.path.exists(support_file) else None,
            'log': f'{prefix}.raxml.log'
        }
    else:
        raise RuntimeError(f"RAxML failed: {result.stderr}")

# Run analysis
results = run_raxml('alignment.fasta', model='GTR+G', bootstraps=100)
print(f"Best tree: {results['best_tree']}")

Model Selection with ModelTest-NG

# Find best model
modeltest-ng -i alignment.fasta -d nt -p 4

# Use selected model in RAxML
raxml-ng --msa alignment.fasta --model <selected_model> --prefix final

Output Files

Tips for Large Datasets

• Use --threads auto for automatic thread detection
• For very large alignments, use --redo to resume interrupted runs
• Consider using site repeats: --site-repeats on
• Use binary MSA format for faster I/O: --parse first