This software is quite old (2011). There are better techniques nowadays. A better de Bruijn graph construction algorithm is BCALM: https://github.com/Malfoy/bcalm (readable Python code here: https://github.com/rchikhi/python-bcalm). Another is an implementation included in the book Genome-Scale Algorithm Design (Chapter 13.2, http://www.genome-scale.info/implementations.html).
Software that constructs the de Bruijn graph of a set of reads (FASTA or FASTQ file). Edges are the (k+1)-mers that appear in the reads, nodes are the k-mers. No distintion is made between a k-mer (resp. (k+1)-mer) and its reverse-complement. It returns a graph in the KisSplice format (ad-hoc) or DOT format (can be opened by most applications, including Zgrviewer and Gephi).
The de Bruijn graph is useful for many next-generation sequencing applications, including de novo genome assembly and variant detection.
For k <= 32, this implementation uses
max( 64*N/p, 64*G )
bits of memory, where:
- N is the number of k-mers present in the reads (N = number of reads * (read length - k + 1))
- G is the number of distinct k-mers in the genome (G = roughly the size of the genome).
- p is the number of passes (specified by option -p in the software).
For 3 billion distinct k-mers, assuming sufficiently many passes, it should construct the graph in 24 Gb of memory.
Note: for 32 <= k <= 64, the constant 64 becomes 128. K-mers larger than 64 nucleotides are not supported.