Contrast-FEL Documentation

Contrast-FEL (Contrast Fixed Effects Likelihood) is a method used to assess whether selective pressures differ between two or more sets of branches at a particular site in a phylogenetic tree. By estimating site-specific nonsynonymous (dN) and synonymous (dS) substitution rates, this method incorporates likelihood ratio tests (LRT) and permutation tests to evaluate the significance of the differences observed.

Citation

If you use Contrast-FEL in your analysis, please cite:

Kosakovsky Pond, Sergei L., et al. "Contrast-FEL—a test for differences in selective pressures at individual sites among clades and sets of branches." Molecular biology and evolution 38.3 (2021): 1184-1198.

Input Parameters

Required Inputs

• Alignment File: An in-frame codon alignment file in a supported format (e.g., .fasta, .phy).
• Branch Sets: A specification of which branches to compare (e.g., "Source" and "Test" groups).

Optional Inputs

• Genetic Code: The genetic code to use for translation (default: "Universal").
• Synonymous Rate Variation: Enable/disable synonymous rate variation (default: "Yes").
• Permutations: Specify whether to perform permutation tests (default: "Yes").
• P-value Threshold: The significance value for site tests (default: 0.05).
• Q-value Threshold: The significance value for False Discovery Rate reporting (default: 0.20).
• Output File: Specify an output file for the results (default is automatic JSON file generation).

Outputs

Summary

Contrast-FEL generates a JSON file that contains:

• Metadata about the analysis and input parameters.
• Site-by-site substitution rate estimates and significance statistics.

Site-Level Output Details

For each codon in the output:

• Alpha (α): Relative synonymous substitution rate.
• Beta (β): Relative nonsynonymous substitution rates for tested branches.
• P-value: Significance of the selection test (lower values indicate stronger evidence of differential selection).
• Q-value: Adjusted p-value for False Discovery Rate.
• Total Branch Length: Length of contributing branches used for inference.
• Substitution Counts: Counts of substitutions mapped to each branch set.

Visualization

Analyses from Contrast-FEL can be visualized using interactive tools available in the web interface. Visualization features include:

• Site-Level Plots: Display of alpha and beta estimates across sites.
• Statistical Summaries: Interactive tables showcasing substitutions and significance levels.
• Phylogenetic Trees: Graphical representation of the tree structure highlighting selected branches and results.

Example Workflow

1. Upload Data:

   • Access the Contrast-FEL webpage and select the relevant alignment with phylogenetic tree file.
   • Define the branch sets you want to compare.

2. Configure Analysis:

   • Choose optional settings such as genetic code.

3. Run Analysis:

   • Click the "Run Analysis" button to initiate the Contrast-FEL computation.

4. Review Results:

   • Upon completion, explore the results via summary statistics, p-values, q-values, and visualizations.
   • Access the detailed JSON output for downloadable insights.

5. Export and Document:

   • Results can be exported for further analysis or reporting purposes.

Example CLI Usage of Contrast-FEL

To use Contrast-FEL for selection analysis, you can run the following command:

/path/to/hyphy/hyphy contrast-fel \
  --alignment <alignment_file> \
  --tree <tree_file> \
  --branch-set <branch_set_1> \
  --branch-set <branch_set_2> \
  --output <output_file> \
  --code <genetic_code> \
  --srv <synonymous_rate_variation> \
  --permutations <perform_permutations> \
  --p-value <p_value_threshold> \
  --q-value <q_value_threshold>

Required Input Parameters

• --alignment: Specify the in-frame codon alignment file (e.g., .fasta, .phy).
• --tree: Provide the path to the phylogenetic tree file.
• --branch-set: Indicate the branches to be used for comparison (repeat for multiple branch sets).

Optional Input Parameters

• --code: Which genetic code should be used (default: "Universal").
• --srv: Include synonymous rate variation in the model ("Yes" or "No", default: "Yes").
• --permutations: Perform permutation significance tests ("Yes" or "No", default: "Yes").
• --p-value: Significance value for site tests (default: 0.05).
• --q-value: Significance value for False Discovery Rate reporting (default: 0.20).
• --output: Specify the output file path for results (default is automatic JSON generation).

Full Example Command

/path/to/hyphy/hyphy contrast-fel \
  --alignment my_alignment.phy \
  --tree my_tree.nwk \
  --branch-set Source \
  --branch-set Test \
  --output results.json \
  --code Universal \
  --srv Yes \
  --permutations Yes \
  --p-value 0.05 \
  --q-value 0.20

Minimal Example Command

/path/to/hyphy/hyphy contrast-fel \
  --alignment my_alignment.phy \
  --tree my_tree.nwk \
  --branch-set Source \
  --branch-set Test \
  --output results.json

FAQ

1. How should I choose test and reference branches when running Contrast-FEL?

The selection of test and reference branches must align with your biological hypothesis. Here are some tips:

• For comparing multiple groups, you can specify multiple branch sets, which generates all pairwise tests.
• Utilize built-in sets like "Internal", "Leaves", and "Unlabeled branches" to focus your analysis on specific parts of your tree.
• For specific comparisons, use annotations in your input tree or selection to ensure you're testing the desired taxa.

2. Why might I see significant differences in dN/dS ratios but no evidence of selection?

Significant differences in dN/dS ratios can occur without evidence of directional selection if the sample sizes for your groups are too small, which leads to imprecise estimates of selection pressure.

3. Can I use Contrast-FEL for analyzing more than two groups?

Yes, Contrast-FEL can conduct analyses on multiple groups. The analysis will generate pairwise comparisons and an omnibus test to assess overall differences among the selected branches.

4. How do I interpret conflicting selection signals from Contrast-FEL and other tests?

It’s essential to recognize that different methods can yield varying results based on their underlying models and assumptions. If Contrast-FEL indicates selection at specific sites while models like BUSTED-PH does not find evidence for selection, consider:

• The context of site-level versus gene-level assessments.
• The robustness of your dataset in terms of size and variability.
• Functional annotations and biological relevancy of the findings might help clarify discrepancies.