hyphy
hyphy copied to clipboard
Interpreting BUSTED-PHenotype analysis
Hi,
I'm exploring the BUSTED-PHenotype analysis. This is the result of it. I didn't quite understand the results.
### Obtaining branch lengths and nucleotide substitution biases under the nucleotide GTR model
>kill-zero-lengths –> Yes
* Log(L) = -13403.35, AIC-c = 27012.97 (103 estimated parameters)
* 1 partition. Total tree length by partition (subs/site) 1.640
### Obtaining the global omega estimate based on relative GTR branch lengths and nucleotide substitution biases
* Log(L) = -12132.66, AIC-c = 24488.27 (111 estimated parameters)
* 1 partition. Total tree length by partition (subs/site) 1.812
* non-synonymous/synonymous rate ratio for *background* = 0.1010
* non-synonymous/synonymous rate ratio for *test* = 0.1408
### Improving branch lengths, nucleotide substitution biases, and global dN/dS ratios under a full codon model
* Log(L) = -12031.98, AIC-c = 24286.91 (111 estimated parameters)
* non-synonymous/synonymous rate ratio for *background* = 0.0807
* non-synonymous/synonymous rate ratio for *test* = 0.1414
### Performing the full (dN/dS > 1 allowed) branch-site model fit
* Log(L) = -11946.53, AIC-c = 24132.14 (119 estimated parameters)
* For *test* branches, the following rate distribution for branch-site combinations was inferred
| Selection mode | dN/dS |Proportion, %| Notes |
|-----------------------------------|---------------|-------------|-----------------------------------|
| Negative selection | 0.000 | 0.237 | |
| Negative selection | 0.022 | 98.329 | |
| Diversifying selection | 13.114 | 1.434 | |
* For *background* branches, the following rate distribution for branch-site combinations was inferred
| Selection mode | dN/dS |Proportion, %| Notes |
|-----------------------------------|---------------|-------------|-----------------------------------|
| Negative selection | 0.054 | 98.476 | |
| Neutral evolution | 1.000 | 1.226 | |
| Diversifying selection | 11.982 | 0.297 | |
### Performing the constrained (dN/dS > 1 not allowed) model fit
* Log(L) = -11960.15, AIC-c = 24157.36 (118 estimated parameters)
* For *test* branches under the null (no dN/dS > 1 model), the following rate distribution for branch-site combinations was inferred
| Selection mode | dN/dS |Proportion, %| Notes |
|-----------------------------------|---------------|-------------|-----------------------------------|
| Negative selection | 0.000 | 88.119 | |
| Negative selection | 0.000 | 0.359 | Collapsed rate class |
| Neutral evolution | 1.000 | 11.522 | |
* For *background* branches under the null (no dN/dS > 1 model), the following rate distribution for branch-site combinations was inferred
| Selection mode | dN/dS |Proportion, %| Notes |
|-----------------------------------|---------------|-------------|-----------------------------------|
| Negative selection | 0.054 | 98.439 | |
| Neutral evolution | 1.000 | 1.269 | |
| Diversifying selection | 12.005 | 0.292 | |
### Performing the constrained background (dN/dS > 1 not allowed on background branches) model fit
* Log(L) = -11961.08, AIC-c = 24159.22 (118 estimated parameters)
* For *test* branches under the null (no dN/dS > 1 on background branches model), the following rate distribution for branch-site combinations was inferred
| Selection mode | dN/dS |Proportion, %| Notes |
|-----------------------------------|---------------|-------------|-----------------------------------|
| Negative selection | 0.000 | 0.213 | |
| Negative selection | 0.021 | 98.349 | |
| Diversifying selection | 12.968 | 1.438 | |
* For *background* branches under the null (no dN/dS > 1 on background branches model), the following rate distribution for branch-site combinations was inferred
| Selection mode | dN/dS |Proportion, %| Notes |
|-----------------------------------|---------------|-------------|-----------------------------------|
| Negative selection | 0.039 | 95.289 | |
| Neutral evolution | 1.000 | 3.626 | |
| Neutral evolution | 1.000 | 1.085 | Collapsed rate class |
### Performing the shared distribution (same on test and background brances) model fit
* Log(L) = -11954.06, AIC-c = 24137.11 (114 estimated parameters)
* For the shared rates model (same between test and background), the following rate distribution for branch-site combinations was inferred
| Selection mode | dN/dS |Proportion, %| Notes |
|-----------------------------------|---------------|-------------|-----------------------------------|
| Negative selection | 0.056 | 98.922 | |
| Neutral evolution | 1.000 | 0.675 | |
| Diversifying selection | 12.466 | 0.402 | |
----
## Branch-site unrestricted statistical test of episodic diversification and association with phenotype/trait [BUSTED-PH]
Likelihood ratio test for episodic diversifying positive selection on test branches , **p = 0.0000**.
Likelihood ratio test for episodic diversifying positive selection on background branches , **p = 0.0000**.
Likelihood ratio test for differences in distributions between **test** and **background** , **p = 0.0101**.
## Analysis summary (p = 0.05)
Selection is acting on the branches with the phenotype / trait, but is **also** acting on background branches.
There is a significant difference between test and background branches in terms of selective pressure
So, it seems like in both groups BUSTED detected a signature of diversifying selection, although there's also between the two there's a significant difference. I don't understand which values should I look at and in which group the positive selection is higher or lower.
Thanks a lot Francesco
Dear @francicco,
Looks like your alignment is subject to selection throughout the entire tree (both test and background branches). If you take point estimates as a rough guide, the test branches have ~1.4% of branch-site combinations evolving with ω ≈ 13, while the background branches have ~0.3% of branch-site combinations evolving with ω ≈ 12. Both of those are significant, statistically (p << 0.001 for both).
However, they are also different from each other, i.e. with p = 0.01 we can reject the hypothesis that ω distributions are the same for background and test branches. BUSTED-PH does not test for directions of change (one is more/less selected) directly, just that they are different. You may use something like the RELAX test to look for such directions (relaxation/intensification).
Does this make more sense?
Best, Sergei
Dear @spond,
Thanks a lot, that makes sense, and yes, I was already thinking to run RELAX to understand the direction! Thanks a lot! F
Stale issue message