BCALM 2 is a bioinformatics tool for constructing the compacted de Bruijn graph from sequencing data.
Read the instructions below to compile, then:
./bcalm -in [reads.fa] -kmer-size [kmer_size] -abundance-min [abundance_threshold]
e.g.
./bcalm -in reads.fastq -kmer-size 21 -abundance-min 2
Importants parameters are:
-kmer-size [int]
The k-mer size, i.e. length of the nodes of the de Bruijn graph.
-abundance-min [int]
Sets a threshold X below which k-mers that are seen (strictly) less than X times in the dataset are filtered out; i.e. sequencing errors, typically.
GCC >= 4.8 or a very recent C++11 capable compiler
Download the latest Linux/MacOS binaries, or compile from source as follows:
git clone --recursive https://github.com/GATB/bcalm
cd bcalm
mkdir build; cd build; cmake ..; make -j 8
You can also install bcalm from bioconda with conda:
conda install -c conda-forge -c bioconda bcalm
File input format can be fasta, fastq, either gzipped or not. BCALM 2 does not care about paired-end information, all given reads contribute to k-mers in the graph (as long as such k-mers pass the abundance threshold).
To pass several files as input:
ls -1 *.fastq > list_reads
./bcalm -in list_reads [..]
BCALM 2 outputs the set of unitigs of the de Bruijn graph. A unitig is the sequence of a non-branching path. Unitigs that are connected by an edge in the graph overlap by exactly (k-1) nucleotides. For a formal description of what BCALM2 outputs, see here
We have two output formats: FASTA and GFA.
GFA output: use scripts/convertToGFA.py to convert the output of BCALM 2 to GFA (contributed by Mayank Pahadia).
FASTA output header:
><id> LN:i:<length> KC:i:<abundance> km:f:<abundance> L:<+/->:<other id>:<+/-> [..]
Where:
-
LNfield is the length of the unitig -
KCandkmfields are for total abundance and mean abundance of kmers inside the unitig, respectively. -
Edges between unitigs are reported as
L:x:y:zentries in the FASTA header (1 entry per edge). A classic forward-forward outcoming edge is labeledL:+:[next node]:+. A forward-reverse,L:+:[next node]:-. Incoming edges are encoded as outcoming edges of the reverse-complement node. E.g.L:-:[previous node]:+means that if you reverse-complemented the current node, then there would be an edge from the last k-mer of current node to the first k-mer of the forward strand of [previous node].
FASTA output header with -all-abundance-counts:
><id> LN:i:<length> ab:Z:<abundance_0> <abundance_(length-k)> L:<+/->:<other id>:<+/-> [..]
Where:
abfield is a vector containing all the kmer abundance in the unitig
BCALM 2 converts all k-mers into their canonical representation with respect to reverse-complements. In other words, a k-mer and its reverse complement are considered to be the same object, appearing only once in the output, either in forward or reverse orientation.
Note: in the output of BCALM 2, each unitig may be either be returned in forward or reverse orientation, with no guarantee that the orientation will stay the same across identical runs of the software.
For a formal description of how BCALM2 handles double-strandedness of DNA, see here
BCALM 2 supports arbitrary large k-mer lengths. You need to recompile it from sources. For k up to, say, 320, type this in the build folder:
rm -Rf CMake* && cmake -DKSIZE_LIST="32 64 96 128 160 192 224 256 320" .. && make -j 8
For compilation, list of kmers should only contain multiples of 32. Also, for technical reason, keep 32 in the list. Of course, for higher k's, BCALM will run slower. Intermediate values create optimized code for smaller KSIZE_LIST="32 320" but then using k values above 32 would be as slow as if k was equal to 320.
After that, BCALM 2 can be run with any k value up to the largest one specified during compilation.
BCALM 2 produces some intermediate files: a .h5 file (or a _gatb/ folder), which contain the k-mer counts. The "*glue*" files contain compacted sequences that needs to be glued together (see BCALM 2 paper). Those files can be safely deleted after an execution, as the actual output is just the FASTA file containing the unitigs.
If using BCALM 2, please cite: Rayan Chikhi, Antoine Limasset and Paul Medvedev, Compacting de Bruijn graphs from sequencing data quickly and in low memory, Proceedings of ISMB 2016, Bioinformatics, 32 (12): i201-i208. (Bibtex)
This project has been supported in part by NSF awards DBI-1356529, CCF-1439057, IIS-1453527, and IIS-1421908.
The legacy, single-threaded code of BCALM version 1 is still available at: https://github.com/Malfoy/bcalm