Biology Reference
In-Depth Information
Table 1
(continued)
Algorithmic
parameter
Software setting
choices
Software setting
Description
This determines whether SAT
´
will be run in parallel mode
Parallelization
“CPU(s) Available”
1-16
Multi-gene
analysis
“Multi-Locus Data”
checkbox/”Sequence
files” button
Checked/unchecked
Folder dialog box
This enables a multi-gene
analysis. See the “Advanced
Analysis” section
Checking this makes SAT
´
perform a RAxML analysis of
the final alignment
Miscellaneous
algorithmic
modifications
“Extra RAxML Search”
checkbox
Checked/unchecked
Miscellaneous
algorithmic
modifications
“Two-Phase
(not SATe)” checkbox
Checked/unchecked
Check to run a two-phase
analysis (first align and then
compute an ML tree)
Choosing one of the settings in the “Quick Set” dropbox will automatically configure the software settings to perform
one of the SAT
´
-II analyses described in ref.
23
. Subsequent modifications to software settings will cause the “Quick Set”
dropbox to display the “(Custom)” choice
5 Additional Guidelines for Selecting Algorithmic Parameters
“Aligner” method
. The choice of method to align the subsets has a
large impact on the resultant alignment and tree. The default is
MAFFT, due to its high accuracy on both simulated and biological
data on both nucleotides and amino acid datasets [
2
,
3
,
13
,
14
,
23
,
24
]. However, Prank has also been used in studies [
24
], and has
the advantage over MAFFT and other standard alignment methods
of not “over-aligning” as much. Because Prank is slower than
MAFFT, the use of Prank to align subsets should be accompanied
by a reduction in the maximum subset size so that the runs can
complete. Finally, Opal and ClustalW are also enabled. Opal pre-
sents memory challenges on large datasets, and is not recom-
mended unless the dataset is small enough. ClustalW is fast and
can be used on any dataset size, but may not provide the same
accuracy as MAFFT.
“Merger” method
. Only Muscle and Opal are enabled for merging
alignments. Muscle is the current default, because it has low mem-
ory requirements while Opal has high memory requirements.
However, we strongly recommend Opal because it generally pro-
duces more accurate alignments. Therefore, we recommend using
Opal unless you do not have sufficient memory for your dataset
analysis. However, this is unlikely to be a problem except for very
large datasets (with more than 10,000 sequences), if you have a
reasonable amount of memory on your laptop or desktop machine.
