-
Notifications
You must be signed in to change notification settings - Fork 5
Analysis run through
A typical SUPERSMART run consists at least of the steps described on this page, in the order in which they are given here. The examples, such as for the primates and the piperaceae follow these same steps.
A typical run consists of a series of subcommands of the smrt program. For example, the first step is the smrt taxize subcommand. Each of these subcommands provides a usage help message that is displayed when issuing either smrt help <subcommand>, or smrt <subcommand> --help.
Each subcommand will also issue a large stream of logging messages, which are colored in the terminal window. Green and blue messages (informational and debugging, respectively) simply inform of the progress of the analysis. Yellow warning messages indicate recoverable issues, though make sure that you understand what the messages mean. Red error messages identify serious, non-recoverable issues.
In the descriptions below, all optional arguments are also described. For the most part these are names for input and output files. Unless you have good reasons it is probably best to go with these defaults because other subcommands downstream also assume them.
The behavior of some commands is additionally influenced by certain variables (in ALL CAPS). These variables can be specified either for your entire SUPERSMART installation, or just for a specific run. To update a variable for your entire installation, run the smrt-config command, which opens an editor to update the configuration file supersmart.ini. To update a variable for a specific run, you can set it as an environment variable with the SUPERSMART_ prefix. For example, to change the number of parallel nodes that a certain command can use to 2 (by default this is NODES=4), do this:
export SUPERSMART_NODES=2
This step takes either a text file with a list of species names, or a higher taxon name provided as a command line argument, and performs taxonomic name resolution. The result is a tab separated spreadsheet where each row represents a species that has been resolved, and the columns give the higher taxa (genus, family, etc.) for this species. To run taxize on a file, do as follows:
smrt taxize -i <filename>
where <filename> is the name of a text file that has on each line a species name to resolve. To run taxize on a higher taxon, do as follows:
smrt taxize -r <taxon>
where <taxon> is a higher taxon name, such as Primates.
Specifies a different name for the output spreadsheet. The default is species.tsv
Specifies to which taxonomic rank to expand the higher taxon. Typically the default, i.e. species, works best. In any case it has to be a taxonomic rank that is actually used by NCBI to annotate individual sequences with, so this could be below species level (e.g. a subspecies) but not above it.
This step creates a Newick tree file that represents the taxonomy for the species spreadsheet. As such, this tree file will have many polytomies (e.g. if more than two genera are in the same family), no branch lengths, and many unbranched internal nodes (e.g. for a monotypic genus there would be a nesting where the genus node has only a single child). This tree is used to generate starting trees.
Specifies a different name for the species spreadsheet. Default is species.tsv
#####
-o <outfile>Specifies a different name for the output tree file. Default isclassification-tree.dnd
This step selects and aligns all phylota clusters that are relevant for the input species. During this step, many FASTA files are produced, as well as an index file called aligned.txt. The FASTA files are named after the phylota cluster they are based on, so the name consists of <seed gi>-<MRCA taxon ID>-<cluster ID>-<cluster type>.fa.
Specifies a different name for the species spreadsheet. Default is species.tsv
Specifies a different name for the output index file. Default is aligned.txt
The alignments from the previous step consists of sets of sequences that have been clustered up to a certain higher taxonomic level (e.g. a genus or a family). It is of course possible that the same marker has been sequenced more broadly (e.g. among multiple genera or families). This step therefore assess which alignments from the previous step are orthologous with one another, and merges them. The result consists of many FASTA files, each named clusterXXX.fa (where XXX is a unique integer), and an index file that lists all the created FASTA files called merged.txt.
To specify the location of the index file from the previous step. Default is aligned.txt
To specify the location of the output index file. Default is merged.txt
This step merges the orthologous clusters into a supermatrix. Candidate clusters to merge into the supermatrix are filtered according to the following variables:
-
BACKBONE_MAX_DISTANCEdefines the maximum average pairwise uncorrected distance across the candidate alignment. If a candidate is more divergent than that it is omitted. Reasonable values are probably in the range of 0.05 to 0.1. -
BACKBONE_MIN_COVERAGEdefines the minimum number of alignments that every species in the supermatrix must participate in. If this value is too high, too many alignments will be concatenated and the supermatrix will become too big to handle in the next steps. Reasonable values are in the range of 3 to 5.
These values can be specified in two ways:
- by editing the configuration file using the
smrt-configcommand. This means that the newly updated value will be used by all subsequent runs. - by setting environment variables, prefixed with
SUPERSMART_(example:SUPERSMART_BACKBONE_MAX_DISTANCE).
Specifies the index file for the orthologous clusters. Default is merged.txt
Specifies the species spreadsheet. Default is species.tsv
Specifies the supermatrix output file name. Default is supermatrix.phy
Specifies names of taxa present in the species spreadsheet that must be included in the supermatrix, regardless their marker coverage. By default, exemplar taxa are selected algorithmically.
Specifies the name for an output spreadsheet summarizing the selected exemplar taxa and markers. Default is markers-backbone.tsv
Constructs a backbone phylogeny using the supermatrix from the previous step. Three inference tools are available: ExaML, ExaBayes and RAxML. For the maximum-likelihood tools, a bootstrap analysis can be specified. The following variables are important:
- For ExaML and RAxML, the substitution model can be specified using the variables
EXAML_MODELorRAXML_MODEL, respectively. These can be specified usingsmrt-configor by setting environment variables with the prefixSUPERSMART_(e.g.SUPERSMART_EXAML_MODEL). - For ExaBayes, the number of runs (
EXABAYES_NUMRUNS), the number of chains (EXABAYES_NUMCHAINS) and the number of generations (EXABAYES_NUMGENS) can be specified.
We have found that the defaults for these variables work reasonably well in most cases.
Specifies the name of the supermatrix. Default is supermatrix.phy
Specifies the name of the classification tree, whose polytomies will be randomly resolved to obtain a starting tree. Default is classification-tree.dnd.
Specifies the name of the inference tool to use. Can be one of examl, raxml or exabayes
Using an ML tool, specifies the number of bootstrap replicates.
Specifies the name of the output tree file. Default is backbone.dnd
Specifies that all intermediate files are to be cleaned up.
The tree(s) from the previous step are unrooted. As we are constructing time-calibrated species trees the trees must be rerooted. This step does this using one of two possible approaches:
- Using an outgroup. This outgroup can be a single species, a comma-separated list of species, or of higher taxa. For example, in the phylogeny of the Primates, the Strepsirrhini (lemurs, bush babies, and such) are the outgroup of the monkeys, apes and tarsiers. Therefore, this higher taxon can be used to root the tree. This approach of course requires some prior knowledge about what would be a good outgroup.
- Using the taxonomy. With this approach, the species spreadsheet is used to identify the rooting that minimizes the number of non-monophyletic higher taxa.
Using either approach, the output is a Newick tree file with the rerooted tree(s) called backbone-rerooted.tsv.
Specifies the species spreadsheet. Default is species.tsv
Specifies the backbone tree file. Default is backbone.dnd
Specifies outgroup taxa.
Specifies the output file. Default is backbone-rerooted.dnd
Smooths the placement of the root along the basal branch. The approach is to select a placement that most approximates the midpoint, such that the average root-to-tip path length is approximately equal for the taxa on both sides of the root.
Time-calibrates the rooted backbone tree(s) using treePL. Calibration points for this step are specified using a tab-separated spreadsheet such as this. You can make such a file for example in Excel or in a text editor, but make sure that:
- The separators really are tab characters, not something else.
- The line breaks are UNIX line breaks, not Windows or MacOS.
The following columns are required, but may be used in any order:
-
NFos- a unique integer for every fossil -
FossilName- a human readable name for the fossil -
CrownvsStem- whether the fossil is astemor acrownfossil -
CalibratedTaxon- which higher taxon the fossil belongs to -
MinAge- minimum age -
MaxAge- maximum age -
BestPracticeScore- currently not used, but refers to the fossil vetting ranking used by FossilCalibrations.org
The location of your fossil spreadsheet must be provided on the command line, so at a minimum the command must be invoked like this:
smrt bbcalibrate -f <fossils>
where <fossils> is the name of the spreadsheet file.
Note: Once the FossilCalibrations.org project makes available either a web service API or a data dump, this step will change accordingly, presumably to a version that provides (semi-)automated fossil harvesting functionality.
The following arguments may optionally be provided, in addition to the mandatory -f <fossils> argument.
Specifies the rerooted backbone tree file. Default is backbone-rerooted.dnd
Specifies the supermatrix file. Default is supermatrix.phy
Specifies the output tree file. Default is chronogram.dnd
This step calculates a consensus tree over a set of Newick trees. This is typically done using the output from the previous bbcalibrate command, but it can equally be used with the output from the preceding commands that produce sets of Newick trees, i.e. bbreroot and bbinfer. The output of this command is a NEXUS file that uses the additional syntax for node annotations that is understood by the highly recommended tree viewer FigTree.
Specifies the input tree file. Default is chronogram.dnd
Specifies the output tree file (NEXUS format). Default is consensus.nex.
Specifies the fraction of burnin to omit. This option necessarily only applies to sets of trees that were generated in Bayesian analyses.
Specifies how to summarize node heights, compatible with syntax for TreeAnnotator, i.e. one of keep, median, mean or ca. Default is ca.
Support threshold above which a node is included in the consensus. Default is 0.0.
Specifies the fossil table file. If provided, the FossilNames are attached to their respective nodes, where possible.
Flag to indicate that the support values are posterior probabilities (and should be embedded in the node annotations as such). If not provided, the support values are assumed to be bootstrap values.
Decomposes the backbone topology into monophyletic clades to be refined in clade-level analyses. For each of these clades, a directory named cladeXXX (where XXX is a unique integer) is created, in which candidate alignments are written. In addition, a tab-separated spreadsheet file - called markers-clades.tsv - is created that summarizes the selected markers. The selection process for these alignments is influenced by two variables:
-
CLADE_MAX_DISTANCEspecifies the maximum average pairwise uncorrected distance below which alignments are still selected (above this value they are assumed to be too saturated to be useful). Reasonable values appear to be around0.1, higher values would select more alignments. -
CLADE_MIN_DENSITYspecifies the minimum fraction of the total species diversity of the clade above which alignments are selected. For example, a value of0.5selects all alignments that have sequence data for at least half of the taxa in the clade.
These variables can be set using smrt-config or as environment variables with the SUPERSMART_ prefix.
Specifies the backbone tree file to decompose. Default is consensus.nex
Attempt to automatically add outgroup from sister genus for each clade, if there is sufficient marker overlap between clades.
Specifies the classification tree used for identifying possible outgroups (i.e. in combination with the -g argument). Default is classification-tree.dnd
Specifies the index file with all candidate alignments. Default is aligned.txt.
Specifies the name for the output tab-separated spreadsheet that summarizes the selected markers. Default is markers-clades.tsv
The clade-level analyses are performed using *BEAST. To this end, the candidate alignments from the previous step must be merged in such a way that it is clear to which species each selected sequence belongs (as there are multiple alignments, and multiple sequences per species in each alignment). As an intermediate format, this merged data is written as richly-annotated NeXML. The maximum number of alignments that is selected is specified with the CLADE_MAX_MARKERS variable: all alignments are sorted by decreasing taxon coverage, and then the first CLADE_MAX_MARKERS are included in the merged data set.
Specifies that the candidate alignments must be enriched with additional haplotypes. The number of haplotypes that is added is specified with the CLADE_MAX_HAPLOTYPES variable.