Here we only describe the parameters adjustable via the ARB interface.
Weighting mask
Specify a weighting mask for the alignment. This increases penalty for mismatches in conservative regions and decreases it in variable regions of the alignment.
Since RAxML only accepts natural numbers as weights, ARB has to multiply the weights of e.g. POS_VAR_BY_PARSIMONY, such that the smallest weight equals 1.
As a consequence the likelihood of the calculated tree is ~ 100000 times higher than w/o weighting mask.
Base tree / Use as constraint tree / Generate random starting tree
Specifying a base tree works different depending on several other parameters.
Generally there are four different possibilities:
-
If you don't select a base tree (i.e. select '????') RAxML generates a starting tree using a Maximum Parsimony algorithm
-
If you additionally set 'Generate random starting tree' to 'Yes' RAxML generates a completely random starting tree. On smaller datasets (around 100-200 taxa) it has been observed that this might sometimes yield topologies of distinct local likelihood maxima which better correspond to empirical expectations.
-
If you select a base tree, RAxML adds all species which are marked but are not in tree to this base tree using Maximum Parsimony. The resulting tree is then optimized using the selected RAxML algorithm.
-
If you set 'Use as constraint tree' to 'Yes' the topology of the given base tree will not be changed, only the position of the added species will be rearranged.
Notes:
-
All species contained in the 'Base tree' have to marked - otherwise RAxML will stop with an error.
Nucleotide Substitution Model / Rate Distribution Model / AA Substitution Model
Please refer to the original documentation for details on Substitution Models
Number of rate categories (DNA GTRCAT only)
This option allows you to specify the number of distinct rate categories, into which the individually optimized rates for each individual site are ?thrown? (Default = 25)
Optimize branches/parameters
Specifies that RAxML shall optimize branches and model parameters on bootstrapped trees as well as print out the optimized likelihood. Note, that this option only makes sense when used with the GTRMIX or GTRGAMMA models (or the respective AA models)!
RAxML algorithm
new rapid hill climbing
RAxML will execute the new (as of version 2.2.1) and significantly faster rapid hill-climbing algorithm
old hill climbing
RAxML will execute the slower old search algorithm of version 2.1.3, this is essentially just for backward compatibility.
optimize input tree
RAxML will optimize the model parameters and branch lengths of the selected 'Base tree' under GTRGAMMA
rapid bootstrap analysis
tell RAxML to conduct a rapid Bootstrap analysis and search for the best-scoring ML tree in one single program run.
Uses the seed specified at 'Random seed'
advanced bootstrap + refinement of BS tree
performs a really thorough standard bootstrap. RAxML will refine the final BS tree under GAMMA and a more exhaustive algorithm.
add new sequences to input tree (MP)
performs just pure stepwise MP addition of new sequences to an incomplete starting tree.
You have to mark all species in tree AND all species which should be added to the tree.
Note: RAxML has a bug in the tree-reader and rejects many trees as unrooted/multifurcated. You can to use 'Tree/Beautify Tree' and select the lowest mode (short branches first) as a workaround.
randomized tree searches (fixed start tree)
will perform several randomized tree searches (as specified at 'Number of runs'), that always start from one fixed starting tree.
Random seed
Used as random seed for 'rapid bootstrap analysis'
Initial rearrangement setting
This allows you to specify an initial rearrangement setting for the initial phase of the search algorithm. If you specify e.g. 10 the pruned subtrees will be inserted up to a distance of 10 nodes away from their original pruning point.
If you don’t specify anything here, a "good" initial rearrangement setting will automatically be determined by RAxML.
Number of runs
Enter a number > 1 to run the selected algorithm multiple times. Specifying e.g. '10' will result in 10 generated trees.
Select ## best trees
If 'Number of runs' is > 1, this specifies how many of the generated tree shall be imported or merge using consense.
The trees with the best likelihood will be selected.
What to do with selected trees?
Import into ARB
All selected trees will be imported into ARB
Create consense tree
Calls consense on all selected trees and imports the resulting consense tree into ARB.
|