BUSTED

BUSTED (Branch-Site Unrestricted Statistical Test for Episodic Diversification) provides a gene-wide (not site-specific) test for positive selection by asking whether a gene has experienced positive selection at at least one site on at least one branch. When running BUSTED, users can either specify a set of foreground branches on which to test for positive selection (remaining branches are designated "background"), or users can test the entire phylogeny for positive selection. In the latter case, the entire tree is effectively treated as foreground, and the test for positive selection considers the entire phylogeny.

For each phylogenetic partition (foreground and background branch sites), BUSTED fits a codon model with three rate classes, constrained as $\omega_1 \leq \omega_2 \leq 1 \leq \omega_3$. As in other methods, BUSTED simultaneously estimates the proportion of sites per partition belonging to each $\omega$ class. This model, used as the alternative model in selection testing, is referred to as the Unconstrained model. BUSTED then tests for positive selection by comparing this model fit to a null model where $\omega_3 = 1$ (i.e. disallowing positive selection) on the foreground branches. This null model is also referred to as the Constrained model. If the null hypothesis is rejected, then there is evidence that at least one site has, at least some of the time, experienced positive selection on the foreground branches. Importantly, a significant result does not mean that the gene evolved under positive selection along the entire foreground.

BUSTED additionally calculates "Evidence Ratios" (ERs) for each site. The ER gives the likelihood ratio (reported on a log-scale) that the alternative model was a better fit to the data compared to the null model. The ER for each site thus provides descriptive information about whether a given site could have evolved under positive selection. The ERs should not be interpreted as statistical evidence for positive selection at individual sites (instead, methods like MEME, FEL, or FUBAR should be used for detecting selection at individual sites).

For each site, two ERs are reported: the Constrained Model ER and the Optimized Null Model ER. The Constrained Model ER calculates the evidence ratio using model parameters inferred from the Constrained model. By contrast, the Optimized Null model ER re-optimizes parameters inferred using the Constrained model for the given site of interest. These optimized parameter values are then used to calculate the site's ER. Again, while these ERs may be helpful descriptors of selection in the data set, they do not provide statistically valid evidence for positive selection at a site.

Citation

If you use BUSTED in your analysis, please cite the following:

Murrell, B et al. "Gene-wide identification of episodic selection." Mol. Biol. Evol. 32, 1365–1371 (2015)

If you use BUSTED-S in your analysis, please cite the following:

Wisotsky, Sadie R., et al. "Synonymous site-to-site substitution rate variation dramatically inflates false positive rates of selection analyses: ignore at your own peril." Molecular Biology and Evolution 37.8 (2020): 2430-2439.

If you use BUSTED-E in your analysis, please cite the following:

Selberg, Avery, et al. "Minus the Error: Estimating dN/dS and Testing for Natural Selection in the Presence of Residual Alignment Errors." bioRxiv (2024): 2024-11.

Input Parameters

Required Inputs

• Alignment File: An in-frame codon alignment file, supported formats include .fasta, .phy, etc.
• Phylogenetic Tree: A phylogenetic tree (with optional branch length annotations) appended to the FASTA file or embedded within the NEXUS file.
• Genetic Code: The intended genetic code for the analysis (default: "Universal").

Optional Inputs

• Branches to Test: Branches from the phylogeny that you wish to investigate for episodic selection (default: "All").
• Synonymous Rate Variation (BUSTED-S): Option to include variations in synonymous substitution rates (options: "Yes", "No", or "Branch-site"; default: "Yes").
• Error Protection (BUSTED-E): To enhance robustness against errors in alignment, this can be enabled or disabled (default: "No").
• Multiple Hits: Support for handling multiple nucleotide substitutions (options: "None", "Double", or "Double+Triple"; default: "None").

Outputs

Summary

BUSTED produces a JSON file summarizing the analysis results.

Output Elements

• Log-Likelihood Ratio Test (LRT): Measures how much better the data fit a model that allows for positive selection compared to one that does not.
• Distribution Estimates: Information on the distribution of omega ratios (dN/dS), indicating rates of nonsynonymous and synonymous substitutions.
• Site-Specific Likelihoods: Evaluations of likelihood at individual codon sites which are stored site-wise.
• Evidence Ratios: Ratio of the likelihoods for models with and without selection, informing the level of support for selection.

Visualization

• Statistical Summaries: Quick access to summary statistics, including evidence ratios and p-values.
• Distribution Plots: Plots detailing the three ω rate classes, sizes reflecting proportion of sites belonging to respective class, and the difference in ω from background to tested branches.

Example Workflow

1. Upload Data:

   • Select and upload your sequence alignment and corresponding phylogenetic tree file.
   • Choose the appropriate parameters for analysis (genetic code, branches to test).

2. Run Analysis:

   • Submit your data by selecting "Run Analysis".

3. Review Results:

   • Access the results in a structured format, including key statistics, plots, and visualizations.
   • Utilize interactive features to explore different selection sites and their classifications.

4. Export Results:

   • Download the JSON results for archiving or further analyses with associated metadata.

Example CLI Usage

To use the BUSTED statistical test for episodic diversification, you can run the following command:

/path/to/hyphy/hyphy /Users/sweaver/Programming/hyphy/hyphy/res/TemplateBatchFiles/SelectionAnalyses/BUSTED.bf \
--code <GENETIC_CODE> \
--alignment <ALIGNMENT_FILE> \
--tree <TREE_FILE> \
--branches <BRANCHES> \
--srv <INCLUDE_SYNONYMOUS_RATE_VARIATION> \
--rates <OMEGA_RATE_CLASSES> \
--syn-rates <SYNONYMOUS_RATE_CLASSES> \
--multiple-hits <MULTIPLE_HITS_OPTION> \
--grid-size <INITIAL_GRID_SIZE> \
--starting-points <INITIAL_GUESS_COUNT> \
--error-sink <ERROR_SINK_OPTION> \
--output <OUTPUT_FILE> \
--save-fit <SAVE_FIT_OPTION>

Parameters

• --code: Specify the genetic code to be used (e.g. "Universal").
• --alignment: Provide the path to an in-frame codon alignment file (required).
• --tree: Specify the path to a phylogenetic tree file (required).
• --branches: Indicate which branches to test (e.g. "All" or "FG").
• --srv: Decide whether to include synonymous rate variation in the model (e.g. "Yes").
• --rates: Set the number of omega rate classes to include in the model (default is 3).
• --syn-rates: The number of synonymous rate classes to include in the model (default is 3).
• --multiple-hits: Specify whether to include support for multiple nucleotide substitutions (options include "None", "Double", or "Double+Triple").
• --grid-size: Define the number of points in the initial distributional guess for likelihood fitting (default is 250).
• --starting-points: The number of initial random guesses to seed rate values optimization (default is 1).
• --error-sink: An advanced experimental setting; include a rate class to capture misalignment artifacts (e.g. "Yes" or "No").
• --output: Specify the path to save the resulting JSON file (default is to save in the same path as the alignment file with the suffix ".BUSTED.json").
• --save-fit: Choose whether to save the BUSTED model fit to a file (default is "/dev/null" to skip saving).

Full Example Usage

Here's a full example with all parameters specified:

/path/to/hyphy/hyphy busted \
--code Universal \
--alignment /path/to/alignment_file.aln \
--tree /path/to/tree_file.tree \
--branches FG \
--srv Yes \
--rates 3 \
--syn-rates 3 \
--multiple-hits Double \
--grid-size 250 \
--starting-points 1 \
--error-sink No \
--output /path/to/results.json \
--save-fit /dev/null

Minimal Example Command

For a minimal example using only the required parameters:

/path/to/hyphy/hyphy busted \
--alignment /path/to/alignment_file.aln \
--tree /path/to/tree_file.tree

FAQs

1. When should I use BUSTED-S over the original BUSTED method?

• Use BUSTED-S when you expect significant variation in synonymous rates across your sequences and want to mitigate potential biases in the detection of positive selection that could arise from ignoring these variations.

2. What is the primary advantage of using BUSTED-E?

• BUSTED-E leverages an error absorption approach which improves the stability of the test and reduces the likelihood of false positives by ensuring that the variances in parameter estimates do not unduly influence the conclusions drawn from the analysis.

3. What is the difference between BUSTED and aBSREL when analyzing selection on a single branch?

• BUSTED tests the entire foreground group simultaneously and is generally more powerful in detecting positive selection across multiple branches. aBSREL focuses on individual branches, allowing for identification of selection events specific to those branches but may lack power with small sample sizes.

4. How do I interpret the results from a BUSTED analysis?

• When interpreting BUSTED results, focus on the p-value and omega values. A p-value < 0.05 indicates evidence for positive selection. However, be cautious with point estimates of ω to avoid misinterpretation. Consider reporting the proportion of lineages under positive selection along with the p-values.

5. Why does BUSTED show significant results while aBSREL does not (or vice versa)?

• Differences can arise due to the way each method calculates selection. BUSTED combines data from all branches in the test, while aBSREL evaluates individual branches. A small number of sequences or branches can significantly impact results.

6. What should I do if Datamonkey mentions “stop codons” or "nucleotides are out of frame"?

• Ensure that your sequence alignment does not have stop codons in non-terminal sequences. Check the alignment to ensure it meets requirements for codon alignment. Remove any stop codons before performing selection analysis.

7. How can I access site-specific results from BUSTED analysis?

• Using the generated JSON file, you can extract the evidence ratios (ER) for individual sites, which indicate potential selection. Look at the corresponding fields in the JSON output that detail the site classification and odds.

8. What does it mean if my analysis shows infinite branch lengths?

• Infinite branch lengths typically indicate saturation or a lack of informative variation in the dataset for that branch. It's advisable to reassess your sequences and ensure appropriate model selection is performed.

9. Can I run BUSTED without specifying a tree file?

• BUSTED requires a phylogenetic tree. However, you may use software to generate a tree from your alignment (e.g., using PhyML) before running the analysis. Datamonkey will generate a Neighbor-Joining tree if no tree is supplied.

10. What should I do if I have negative results from BUSTED for a gene known to be under positive selection?

• Consider re-evaluating your alignment for inconsistencies, increasing the number of branches in your foreground definition, or combining it with additional analyses (like MEME or FEL) that might provide more targeted site-level insights.

References

• BUSTED-E
• BUSTED-S
• BUSTED