A rarefaction software written in C++11 to rarefy large high count datasets quickly and return diversity measures.

The easiest way is to install from Miniconda (see rtk package) using:

    conda install -c bioconda rtk

For use in R, use install.packages("rtk").

Otherwise, download RTK from https://github.com/hildebra/Rarefaction/releases or compile from source.

To build this software you will need to have a compiler for C++11 on your system. On a GNU/Linux system you usually have to install developer tools to do that. For Ubuntu this is explained here: https://help.ubuntu.com/community/InstallingCompilers

RTK was tested to compile successfully on windows (Microsoft Visual Studio C++ 2017 RC, Windows 10), GNU/Linux (g++ v. 4.8.5 and v. 6.1.1) and on Mac OS 10.11.2 (Apple LLVM version 7.0.0 (clang-700.0.72)).

Compile in UNIX

git clone https://github.com/hildebra/Rarefaction
cd Rarefaction/rtk
make

Two modes for rarefaction of a count table are available

rtk  <mode> -i <input.csv> -o <output> [options]

<mode>  mode can be either swap or memory for rarefaction or 
        colsums for column sums report.
        Swap mode creates temporary files but uses less memory.
        The speed of both modes is comparable.

-i      path to an .csv file to rarefy
-o      path to a output directory
-d      Depth to rarefy to, may be comma seperated list. Default is 0.95 times the minimal column sum.
-r      Number of times to create diversity measures. Default is 10.
-w      Number of rarefied tables to write.
-t      Number of threads to use. Default: 1
-ns     If set, no temporary files will be used when writing rarefaction tables to disk (no swap).

median_alpha_diversity.tsv

This file contains the median diversity measures for all Samples in a tab separated format.

richness|eveness|...|.tsv

Each diversity measures is exported as a table containing all repeats for all sample.

global_diversity.tsv

Holds the ACE,ICE and chao2 for the table.

rarefied_to_X_n_Y.tsv

If NoOfMatrices > 0 each rarefied matrix will be saved in the output directory under this file. The structure of all files is the same and similar to the input file.

sums.txt

This file contains the column sums of all samples. It can be used to estimate well suited rarefaction depth.

If the mode memory is used, temporary files will be produced to reduce RAM usage. Thus the input matrix will be first split into its columns and each column will be written into a single file. Those file will then be loaded again and deleted after the software is finished using them.

Temporary files will also be created if -w > 0. In this case the vectors of the rarefied tables will be stored on disk as binary before merging them to tables. Thi can be prevented by using the -ns flag.

In both cases RAM usage is drastically reduced and the load on the local drive is substantially higher.

Knowing the dataset at hand is relevant. That is why RTK allows the user to quickly estimate the column sums of the dataset.

The mode colsums creates two files containing sorted and unsorted column sums of all samples:

rtk  colsums -i <input.csv> -o <output> [options]

Input data for RTK should be a count table in a .tsv or .csv format. Row and column names must be provided and be unique.

Example file:

Rarefaction is always performed on the columns of the dataset. If you want to rarefy on the rows please consider transposing your input data ahead of rarefaction.

On a UNIX system use AWK to transpose a .csv table: https://stackoverflow.com/questions/1729824/transpose-a-file-in-bash

A minimal working example of a rarefaction is shown here. This example should run on any UNIX system.

#!/bin/bash
FILE="example.input.csv"
touch $FILE
echo -e "OUT    \tSample 1\tSample 2\tSample 3"       >> $FILE
echo -e "OTU 1\t  232      \t  10       \t  0"        >> $FILE
echo -e "OTU 2\t  0        \t  57       \t  22"       >> $FILE
echo -e "OTU 3\t  17       \t  0        \t  45"       >> $FILE
echo -e "OTU 4\t  5        \t  83       \t  0"        >> $FILE

./rtk memory -i $FILE -o test.
ls -lh test.*

If you use RTK in a publication, please consider citing the Bioinformatics application note at: https://academic.oup.com/bioinformatics/article/3111845/RTK-efficient-rarefaction-analysis-of-large

Saary, Paul, et al. "RTK: efficient rarefaction analysis of large datasets." Bioinformatics (2017)

RTK is licensed under the GPLv2. See notice and license file for more information.

Copyright (c) 2016 by Falk Hildebrand and Paul Saary