This glossary provides definitions for terms in the HyPhy-Vision reports for each method. See here for descriptions of each method.

Shared Terms#

• log L - The log likelihood estimate of the respective model fit.
• # par. - The number of estimated parameters in the respective model.
• Time to fit - Total clock time to fit the respective model.
• AIC_c - Small-sample Akaike Information Criterion score.
• L_tree - The tree length under the respective model, where tree length represents the expected number of substitutions per site.

aBSREL Glossary#

Summary: Tree Table#

• $\boldsymbol \omega$ rate classes - Number of $\omega$ rate classes inferred.
• # of branches - The number of branches inferred to have the respective number of $\omega$ rate classes.
• % of branches - The proportion of branches inferred to have the respective number of $\omega$ rate classes.
• % of tree length - The percentage of the total tree length inferred to have the respective number of $\omega$ rate classes.
• # under selection - The number of branches inferred to have undergone positive selection (a rate class of $\omega>1$) at the designated p-value threshold, after correction for multiple testing.

Summary: Model Fits Table#

• MG94 - The baseline model fit of MG94xREV that infers a single $\omega$ per branch.
• Full Model - The full aBSREL model fit to the tree, where the number of $\omega$ classes per branch is inferred adaptively.
• See these Shared Terms for additional reported information.

Full Table#

• Name - Branch of interest, where bolded rows indicate that the branch shows evidence, at the designated p-value threshold, of positive selection.
• B - Optimized (under the full aBSREL model) branch length for the branch of interest
• LRT - Likelihood Ratio Test statistic for selection. This LRT was calculated by comparing the fitted aBSREL model to the null model where rate classes of $\omega>1$ are disallowed.
• Test p-value - P-value corrected for multiple testing using the Holm-Bonferroni correction. These values are only calculated for branches with uncorrected p-values less than 1.
• Uncorrected p-value - Raw p-value before correction for multiple testing.
• $\boldsymbol \omega$ distribution over sites - Inferred $\omega$ estimates and respective proportion of sites along the respective branch.

RELAX Glossary#

Summary: Model Fits Table#

• Null - This model represents the null model used to test for selection relaxation. Here, the selection intensity parameter k is set to 1 for both branch sets (reference and test).
• Alternative - This model represents the alternative model used to test for selection relaxation. Here, the selection intensity parameter k is allowed for vary on the "test" partition of branches. Note that a single k is inferred and shared across all test branches.
• Partitioned MG94xREV - This baseline model fits a single $\omega$ value to each of the two branch sets (reference and test), respectively.
• Partitioned Descriptive - This model infers three $\omega$ classes distributions for each partition, respectively, without using the selection intensity parameter k (e.g. k=1 throughout).
• General Descriptive - This model fits three $\omega$ classes to the entire phylogeny, i.e. shared across all branches. This model then infers a single selection intensity parameter k for each branch.
• Branch set - The set of branches (all, test, or reference) for which the given set of $\omega$ rate classes were inferred.
• $\boldsymbol{\omega_{<1,2,3>}}$ - The inferred $\omega$ value for the respective rate class ($\omega_1$, $\omega_2$, or $\omega_3$). The value in parentheses indicates the proportion of sites inferred to belong to this rate class.

BUSTED Glossary#

Summary: Model Fits Table#

• Constrained model - This model represents the null model used in the BUSTED hypothesis test. For this model, the background and foreground branch partitions share all $\omega$ rate classes, but the value $\omega$ rate class used to test for selection is constrained to equal 1 (i.e. $\omega_3 = 1$ for both the background and foreground branches).
• Unconstrained model - This model represents the alternative model used in the BUSTED hypothesis test. For this model, the value for $\omega$ rate class used to test for selection is permitted to exceed 1 on the foreground branches. (i.e. $\omega_3 > 1$ on foreground branches).
• $\boldsymbol{\omega_{<1,2,3>}}$ - The inferred $\omega$ value for the respective rate class ($\omega_1$, $\omega_2$, or $\omega_3$). The value in parentheses indicates the proportion of sites inferred to belong to this rate class.
• See these Shared Terms for additional reported information.

Summary: Model Evidence Ratios Per Site Table#

• Site Index - The codon site of interest.
• Unconstrained Likelihood - The log likelihood score for the codon site of interest, calculated from the Unconstrained Model fit.
• Constrained Likelihood - The log likelihood score for the codon site of interest, calculated from the Constrained Model fit.
• Optimized Null Likelihood - The log likelihood score for the codon site of interest, calculated from the Optimized Null Model fit. The Optimized Null model is the Constrained model whose parameters have been re-optimized for this specific site.
• Constrained Evidence Ratio - Evidence ratio for positive selection at the given codon site, using the Constrained Model as the null model.
• Optimized Null Evidence Ratio - Evidence ratio for selection at the given codon site, using the Optimized Null model as the null model.