abismal

Another Bisulfite Mapping Algorithm (abismal) is a read mapping program for bisulfite sequencing in DNA methylation studies.

Download the latest stable release here.

See how to get started and the full program documentation.

Requirements

Currently abismal requires a C++ compiler that supports the C++11 standard and OpenMP. The default compiler assumed is g++ (comes with GCC, available on your Linux or macOS machine). The g++ compiler has supported the C++11 standard since roughly 2012 (GCC 4.7) so this should not cause any problems. It also requires an OMP library and headers to be available, which rarely causes problems. Instructions to get HTSlib, for macOS or Linux systems, can be found below.

If you have trouble with the make part of the installation procedure described below, please contact us via e-mail or through a GitHub issue.

Documentation

The full documentation for abismal can be found here. This explains the use of each parameter in full detail. Below, after installation instructions, we describe the most common use cases: indexing a genome and mapping single-end and paired-end reads.

Installation on Linux

These instructions are for building abismal from source, rather than obtaining it through a package manager like conda.

These instructions assume you have access to apt which is installed on Ubuntu-based and Debian-based distributions. The only difference for other linux distributions is how you get the dependencies. Likely all you need is:

$ sudo apt-get install -y libhts-dev

If you don't have adminstrator privileges, there are other options. If you have the libhts-dev installed, to build abismal the following should work:

$ wget https://github.com/smithlabcode/abismal/releases/download/v3.2.4/abismal-3.2.4.tar.gz
$ tar -zxvf abismal-3.2.4.tar.gz
$ cd abismal-3.2.4
$ mkdir build && cd build
$ ../configure --prefix=/where/you/want/abismal
$ make
$ make install

Be sure that you have permissions to write files to /where/you/want/abismal. This will install abismal, abismalidx and simreads inside the bin directory of the specified location.

Installation on macOS

The GitHub repo for abismal includes tests that run on macOS 13 (Ventura), and we use the following steps. Although our tests begin with a "fresh" macOS installation, they have certain tools already available. In particular, Homebrew is already available and possibly some other tools. Homebrew is necessary as the first step to get the tools and dependencies:

$ brew update
$ brew install gcc
$ brew install htslib gsl
$ brew list --versions gcc

At this point, keep the version of gcc in mind, because it will probably be needed below. If you don't already have abismal downloaded, the next step is to download it. Here we will assume you are using a release rather than a clone. To build from a clone involves at least one more step.

$ wget https://github.com/smithlabcode/abismal/releases/download/v3.2.4/abismal-3.2.4.tar.gz
$ tar -zxvf abismal-3.2.4.tar.gz
$ cd abismal-3.2.4

Finally, these steps build the software:

$ mkdir build && build
$ ../configure \
    --prefix=/path/to/install \
    CXX="g++-13" \
    CPPFLAGS="-I$(brew --prefix)/include" \
    LDFLAGS="-L$(brew --prefix)/lib"
$ make
$ make install

Notice the g++-13 in the ../configure command. This is the version number referenced above. If you have a different version number (e.g., when gcc-14 is the default), you will need to update that number to correspond to the major version number. Be sure you have permissions to write to the directory /path/to/install.

How to get the dependencies through conda

If you are on linux and do not have adminstrator privileges to get the dependencies (e.g., HTSlib), you can get them either by building them directly from source, or through conda. In particular, for obtaining HTSlib through conda, do the following:

$ conda install -c bioconda htslib

as explained here at htslib. I used conda obtained through miniconda3, which means the default location for HTSlib to be installed is ~/miniconda3 and then inside the lib and include subdirectores. So once this is done, you can build abismal by replacing the configure step in the earlier explanations by

../configure --prefix=/path/to/install \
    CPPFLAGS="-I${HOME}/miniconda3/include" \
    LDFLAGS="-L${HOME}/miniconda3/lib"

Note that you can use this approach with both Linux or macOS, but in the case of macOS you can replace the LDFLAGS and CPPFLAGS for conda, but keep the CXX variable. Remember not to use tilde (~) in place of the ${HOME} variable above. It might work, but shouldn't.

Installation from a clone of the repo

This method is likely only useful if you need the most recent update, and is not recommended for most users. The only difference from the above explanations for linux and macos is that you will need to clone the repo, which means you need git installed, and you will also need to build the sources in place and without much reporting in case of any problems.

$ cd /where/you_want/the_code
$ git clone --recursive git@github.com:smithlabcode/abismal.git
$ cd abismal
$ make
$ make install

If you are building from the source in a cloned repo, you will likely see other ways to accomplish it by examining the files in the root of the repo.

Indexing the genome

The index can be constructed as follows, based on a genome existing entirely in a single FASTA format file:

$ abismalidx <genome.fa> <index-file>

Bisulfite mapping

single-end reads

$ abismal [options] -i <index-file> -o <output-file> <reads.fq>

paired-end reads

$ abismal [options] -i <index-file> -o <output-file> <read_1.fq> <read_2.fq>

abismal options

option	long version	arg type	default	description
-i	-index	string		genome index file
-g	-genome	string		genome file (FASTA)
-o	-outfile	string		output file (default SAM format)
-s	-stats	string		mapping statistics output file (YAML)
-c	-max-candidates	integer	100	max candidates per seed*
-l	-min-frag	integer	32	minimum fragment length (PE mode)
-L	-max-frag	integer	3000	maximum fragment length (PE mode)
-m	-max-distance	double	0.1	max relative number of errors
-a	-ambig	boolean		report a position for ambiguous reads
-P	-pbat	boolean		input follows the PBAT protocol
-R	-random-pbat	boolean		input follows the random PBAT protocol
-A	-a-rich	boolean		reads are A-rich (SE mode)
-t	-threads	integer	1	number of mapping threads
-v	-verbose	boolean		print more run info
-B	-bam	boolean	output SAM format	write output in BAM format

* the max candidates parameter controls the amount of "effort" in mapping. In the "sensitive" step, which aligns reads with smaller exact match seeds, abismal skips seeds that retrieves more than c candidates. The higher the value of c, the more alignments abismal performs. Note that abismal still aligns reads to every exact match hit that spans more than half of the read ("specific step"). The specific step does not change with the value set by c.

Examples

(1) Indexing the genome

To make an index for hg38:

$ abismalidx hg38.fa hg38.abismalidx

In the process of building the index, the names of chromosomes will be truncated at the first whitespace character.

(2) Bisulfite Mapping

To map single-end reads in file reads.fq to human genome hg38:

$ abismal -i hg38.abismalidx -o reads.sam reads.fq

To map paired-end reads in files reads-1.fq and reads-2.fq to human genome hg38:

$ abismal -i hg38.abismalidx -o reads.sam reads-1.fq reads-2.fq

To map reads in BAM format:

$ abismal -B -i hg38.abismalidx -o reads.bam reads.fq

To map reads to human genome without requiring a separate index file (i.e. run both indexing and mapping simultaneously):

$ abismal -g hg38.fa -o reads.sam reads.fq

Mapping results are reported in SAM format. Some choices in the output are explicitly highlighted below:

• Reads are output identically to how they were read, regardless of mapped strand.
• the NM tag reports the edit distance between the read and the output, specifically the sum of mismatches, insertions and deletions to the best mapping position.
• The CV tag reports the assumed bisulfite base used to map the read. Reads mapped as A-rich will be reported with CV:A:A, and reads mapped as T-rich will be reported with CV:A:T. This tag is independent of the strand the read was mapped to. If reads are not mapped in PBAT or random PBAT mode, the first end will always be T-rich and the second end will always be A-rich.

Contacts

Andrew D Smith andrewds@usc.edu Guilherme Sena desenabr@usc.edu

Citation

The abismal manuscript is available here. If you used abismal to analyze your data, please cite us as follows.

de Sena Brandine, G., & Smith, A. D. (2021).
Fast and memory-efficient mapping of short bisulfite sequencing reads using a two-letter alphabet.
NAR Genomics and Bioinformatics, 3(4), lqab115.

Copyright

Copyright (C) 2018-2023 Andrew D. Smith and Guilherme de Sena Brandine

Authors: Andrew D. Smith and Guilherme de Sena Brandine

abismal is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

abismal is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.