Skip to content

Using Alignment Uncertainty from Many Multiple Sequence Alignments to Create Consensus Trees

Notifications You must be signed in to change notification settings

evanfloden/concTree

Repository files navigation

concTree-NF

Phylogenetic Trees and Bootstrap Support Values Utilising Alignment Uncertainty from Multiple MSA Generated Trees

Motivation

concTree incorporates the information from uncertainty in multiple sequence alignments into the construction of multiple phylogenetic trees and the bootstrap support values.

Schematic Outline

Quick start

Make sure you have all the required dependencies listed in the last section.

Install the Nextflow runtime by running the following command:

$ curl -fsSL get.nextflow.io | bash

When done, you can launch the pipeline execution by entering the command shown below:

$ nextflow run skptic/concTree-nf

By default the pipeline is executed against the provided example dataset. Check the Pipeline parameters section below to see how enter your data on the program command line.

Pipeline parameters

--input

  • Specifies the location of the fasta file(s) containing sequences.
  • Multiple files can be specified using the usual wildcards (*, ?), in this case make sure to surround the parameter string value by single quote characters (see the example below)
  • By default it is set to the concTree-NF's location: ./tutorial/*.fa

Example:

$ nextflow run skptic/concTree-nf --input '/home/dataset/*.fasta'

This will handle each fasta file as a seperate alignment/tree/bootstrap support

--alignments

  • Specifies the number of alignments to sample from the guidance2 alternative alignments.
  • Min = 1, Max = 400
  • By default it is set to 100

Example:

$ nextflow run skptic/concTree-nf --alignments '25'

--output

  • Specifies the folder where the results will be stored for the user.
  • It does not matter if the folder does not exist.
  • By default is set to concTree-NF's folder: ./tutorial/results

Example:

$ nextflow run skptic/concTree-nf --output /home/user/my_results 

Cluster support

concTree-NF execution relies on Nextflow framework which provides an abstraction between the pipeline functional logic and the underlying processing system.

Thus it is possible to execute it on your computer or any cluster resource manager without modifying it.

Currently the following platforms are supported:

  • Oracle/Univa/Open Grid Engine (SGE)
  • Platform LSF
  • SLURM
  • PBS/Torque

By default the pipeline is parallelized by spanning multiple threads in the machine where the script is launched.

To submit the execution to a SGE cluster create a file named nextflow.config, in the directory where the pipeline is going to be launched, with the following content:

process {
  executor='sge'
  queue='<your queue name>'
}

In doing that, tasks will be executed through the qsub SGE command, and so your pipeline will behave like any other SGE job script, with the benefit that Nextflow will automatically and transparently manage the tasks synchronisation, file(s) staging/un-staging, etc.

Alternatively the same declaration can be defined in the file $HOME/.nextflow/config.

To lean more about the avaible settings and the configuration file read the Nextflow documentation.

Dependencies

  • Java 7+
  • Guidance2
  • ClustalW2
  • MAFFT
  • PRANK
  • T-Coffee
  • RAXML

About

Using Alignment Uncertainty from Many Multiple Sequence Alignments to Create Consensus Trees

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published