IQ-TREE Guide for DNA Barcoding

What is IQ-TREE?

IQ-TREE (Efficient and Effective Stochastic Algorithm for Estimating Maximum-Likelihood Phylogenies) is a software tool that builds evolutionary trees (phylogenetic trees) from DNA sequences.

Think of it as a tool that:

Compares DNA sequences from different organisms
Figures out which organisms are most closely related
Creates a family tree showing their evolutionary relationships

Why Do We Use IQ-TREE?

IQ-TREE is the gold standard for phylogenetic analysis in 2025 because:

✓ Fast - Analyzes sequences much faster than older methods ✓ Accurate - Uses advanced statistical models to find the best tree ✓ Automatic - Picks the best evolutionary model for your data ✓ Reliable - Provides confidence scores (bootstrap values) for branches ✓ Widely used - Standard tool in scientific research

Alternatives We Considered

RAxML - Older, slower, but still good
FastTree - Very fast but less accurate
MEGA - GUI-based, good for teaching but less powerful
MrBayes - Uses different method (Bayesian), very slow

We chose IQ-TREE because it gives the best balance of speed, accuracy, and ease of use for students.

What is Maximum Likelihood (ML)?

Maximum Likelihood is a statistical method for building phylogenetic trees. It answers the question:

"Given my DNA sequences, which tree arrangement is most likely to have produced these sequences through evolution?"

Simple Explanation

Imagine you have DNA from 4 mosquito species. There are many ways to arrange them into a tree:

Tree 1:          Tree 2:          Tree 3:
  ┌─A              ┌─A              ┌─B
──┤               ──┤               ──┤
  │  ┌─B            │  ┌─C            │  ┌─A
  └──┤              └──┤              └──┤
     │  ┌─C            │  ┌─B            │  ┌─C
     └──┤              └──┤              └──┤
        └─D              └─D              └─D

Maximum Likelihood asks: "Which tree is most likely to be correct, given:

The DNA differences we see
A model of how DNA evolves (mutations, deletions, insertions)
The laws of probability"

For each tree, ML calculates a likelihood score - how probable that tree is given your data. The tree with the highest score "wins."

Why Maximum Likelihood Instead of Other Methods?

Method	How It Works	Pros	Cons	When to Use
Maximum Likelihood (ML)	Finds tree with highest probability given data	Most accurate; uses statistical model; handles complex evolution	Slower than simpler methods	Best for research & publication
Neighbor-Joining (NJ)	Joins closest sequences first	Very fast; simple	Less accurate; no statistical model	Quick previews, large datasets
Maximum Parsimony (MP)	Finds tree with fewest mutations	Fast; intuitive	Assumes all changes equally likely (wrong!)	Historical; rarely used now
Bayesian (MrBayes)	Samples many trees using probability	Most rigorous statistically	VERY slow (hours to days)	Research with critical questions

For DNA barcoding, Maximum Likelihood is the gold standard because:

✅ Accurate enough for species identification
✅ Fast enough for student projects (1-5 minutes)
✅ Uses a realistic model of DNA evolution
✅ Provides statistical support (bootstrap values)
✅ Widely accepted in scientific literature

Why Not Bayesian?

Bayesian methods (like MrBayes) are theoretically superior BUT:

Time: Takes hours to days for 50 sequences
Complexity: Students must understand priors, MCMC, convergence
Overkill: For species ID, ML is plenty accurate

Our choice: ML gives 95% of the accuracy in 1% of the time.

In Practice

When IQ-TREE runs, you'll see:

Choosing best model... GTR+F+I+G4
Computing ML tree... done
Running 1000 bootstrap replicates... done

This means:

Best model chosen: GTR+F+I+G4 (complex model of DNA evolution)
ML tree computed: Found the most likely tree arrangement
Bootstrap: Tested how confident we are in each branch

How IQ-TREE Works in Our Pipeline

Input

IQ-TREE receives an aligned FASTA file from the alignment step:

>AT99_F
ATCGATCGATCG---ATCG
>AT99_R
ATCGATCGATCG---ATCG
>AT_ROCK_F
ATCGAT---ATCGATGATCG
>AT_ROCK_R
ATCGAT---ATCGATGATCG

All sequences must be the same length (gaps filled with -).

What IQ-TREE Does

Step 1: Choose the Best Model

DNA evolves in different ways (some positions change faster than others)
IQ-TREE tests many models and picks the best one
Common models: GTR+G, HKY+I, K2P

Step 2: Build the Tree

Tries millions of possible tree arrangements
Uses maximum likelihood to find the most probable tree
This is like finding the family tree that best explains the DNA differences

Step 3: Test the Tree (Bootstrap)

Creates 1000 "fake" datasets by resampling your data
Builds a tree for each fake dataset
Counts how often each branch appears
This gives confidence scores (0-100%)

Output

IQ-TREE creates several files:

results/phylogeny/
├── tree.treefile          # The best tree (Newick format)
├── tree.iqtree           # Detailed analysis report
├── tree.log              # What IQ-TREE did step-by-step
├── tree.png              # Simple tree visualization
└── phylogeny_report.html # Interactive HTML report

Understanding the Tree

Newick Format (tree.treefile)

((AT99_F:0.0123,AT99_R:0.0098)100:0.0456,(AT_ROCK_F:0.0234,AT_ROCK_R:0.0221)95:0.0567);

Reading this:

() = groups organisms together
:0.0123 = branch length (amount of evolution)
100 or 95 = bootstrap support (% confidence)

Tree Visualization

              ┌─ AT99_F
    ┌─100%───┤
    │         └─ AT99_R
────┤
    │         ┌─ AT_ROCK_F
    └─95%────┤
              └─ AT_ROCK_R

Interpreting:

Closer together = more related (e.g., AT99_F and AT99_R are very similar)
Bootstrap 100% = we're very confident in this grouping
Bootstrap <70% = less confident, might not be real relationship
Longer branches = more mutations = more evolution

What the Numbers Mean

Branch Length

0.001 = very few mutations (0.1% of positions different)
0.01 = moderate mutations (1% of positions different)
0.1 = many mutations (10% of positions different)

In DNA barcoding (COI gene):

Same species: usually 0-2% difference (0.000-0.02)
Different species, same genus: 2-10% difference (0.02-0.10)
Different genera: >10% difference (>0.10)

Bootstrap Support

90-100% = Strong support - very confident
70-89% = Moderate support - probably correct
50-69% = Weak support - uncertain
<50% = Very weak - might be wrong

Rule of thumb: Only trust branches with >70% bootstrap support.

Reading the IQ-TREE Report (tree.iqtree)

When you open tree.iqtree, you'll see:

1. Model Selection

Best-fit model: GTR+F+I+G4

GTR = General Time Reversible (allows different mutation rates)
+F = Empirical base frequencies (uses your data's A/T/G/C ratios)
+I = Invariable sites (some positions never change)
+G4 = Gamma rate heterogeneity (some positions change faster)

What this means: IQ-TREE chose a complex model that accounts for how DNA really evolves.

2. Likelihood Score

Log-likelihood: -2345.678

Higher (less negative) is better. Only useful for comparing trees from the same data.

3. Tree Statistics

Number of taxa: 52
Number of free parameters: 125

More sequences = more complex tree = needs more data to be confident.

4. Bootstrap Results

Branches with support >70%: 45/50 (90%)

If most branches have high support, your tree is reliable!

Common Questions

Q: Why does my tree look weird?

A: Check if:

You have enough sequences (need at least 4, ideally 10+)
Sequences are actually related (all from the same gene region)
Alignment is good (run quality control first)
You included reference sequences for comparison

Q: What if bootstrap values are low (<70%)?

A: This means:

Not enough data to be confident
Sequences might be too similar (or too different)
Need more sequences to resolve relationships

Solution: Add more reference sequences from GenBank.

Q: Can I use the tree without reference sequences?

A: Yes, but it's limited:

Shows relationships among YOUR sequences
Can't identify species (need references for that)
Can spot contamination (if one sequence is very different)

Q: How long does IQ-TREE take?

4-10 sequences: ~10 seconds
50 sequences: ~1 minute
100 sequences: ~5 minutes
1000 sequences: ~1 hour

Our pipeline uses 1000 ultrafast bootstrap replicates, which is much faster than traditional bootstrap (1000x replicates would take days!).

Viewing Your Tree

Option 1: FigTree (Recommended)

Download free from: http://tree.bio.ed.ac.uk/software/figtree/

Steps:

Open FigTree
File → Open → Select tree.treefile
Check "Node Labels" → "label" to see bootstrap values
Adjust appearance, colors, labels
Export as PDF or PNG

Option 2: iTOL (Web-based)

Upload to: https://itol.embl.de/

Steps:

Upload tree.treefile
Customize colors and labels
Download publication-quality figure

Option 3: Our HTML Report

Open phylogeny_report.html in your browser - includes basic visualization.

Example: Interpreting Your Results

Let's say you sequenced mosquito samples and got this tree:

              ┌─ Your_Sample_1
    ┌─98%────┤
    │         └─ Aedes_aegypti_reference
────┤
    │         ┌─ Your_Sample_2
    └─95%────┤
              └─ Culex_quinquefasciatus_reference

What this tells you:

Your_Sample_1 groups with Aedes aegypti (98% confidence) → Probably Aedes aegypti
Your_Sample_2 groups with Culex quinquefasciatus (95% confidence) → Probably Culex quinquefasciatus
These two mosquitoes are from different genera → Confirms they're very different species

Troubleshooting

Error: "Alignment contains no parsimony-informative sites"

Problem: All sequences are identical (or only differ by gaps) Solution: Check your alignment - you need sequences that actually differ

Error: "Too few sequences"

Problem: IQ-TREE needs at least 4 sequences Solution: Add more sequences or reference sequences

Warning: "Some branch lengths are very long"

Problem: One sequence is very different from others Solution:

Might be contamination (check your sample)
Might be from a different gene region
Might be a different organism entirely

Low bootstrap support

Problem: Can't confidently determine relationships Solution:

Add more sequences
Use longer DNA sequences
Check alignment quality
Make sure sequences are actually related

Summary

What you need to know:

IQ-TREE builds evolutionary trees from DNA sequences
It automatically picks the best method for your data
Bootstrap values show confidence (>70% is good)
Branch lengths show how different sequences are
Use FigTree or iTOL to make pretty trees
Always include reference sequences for species ID

What you DON'T need to worry about:

Complex mathematical models (IQ-TREE handles this)
Manual parameter tuning (automatic mode works great)
Programming (our pipeline runs it for you)

Just understand:

What the tree shows (relationships)
What the numbers mean (confidence and distance)
How to interpret your results (which sequences are related)

Happy tree building! 🌳

FilesExpand file tree

iqtree_guide.md

Latest commit

History