From Indexing Data Structures to de Bruijn Graphs

Abstract : New technologies have tremendously increased sequencing throughput compared to traditional techniques, thereby complicating DNA assembly. Hence, assembly programs resort to de Bruijn graphs (dBG) of $k$-mers of short reads to compute a set of long contigs, each being a putative segment of the sequenced molecule. Other types of DNA sequence analysis, as well as preprocessing of the reads for assembly, use classical data structures to index all substrings of the reads. It is thus interesting to exhibit algorithms that directly build a de Bruijn graph of order $k$ from a pre-existing index, and especially a contracted version of the de Bruijn graph, where non branching paths are condensed into single nodes. Here, we formalise the relationship between suffix trees/arrays and dBGs, and exhibit linear time algorithms for constructing the full or contracted de Bruijn graphs. Finally, we provide hints explaining why this bridge between indexes and dBGs enables to dynamically update the order $k$ of the graph.
Document type :
Reports
[Research Report] RR-14004, LIRMM. 2014


http://hal-lirmm.ccsd.cnrs.fr/lirmm-00950983
Contributor : Eric Rivals <>
Submitted on : Wednesday, March 5, 2014 - 7:31:18 PM
Last modification on : Wednesday, January 20, 2016 - 2:54:09 PM
Document(s) archivé(s) le : Thursday, June 5, 2014 - 12:00:27 PM

File

STSA-DBG-RR.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : lirmm-00950983, version 2

Collections

CIRAD | INRIA | LITIS | IBC | INRA | MAB | IRD | LIRMM

Citation

Bastien Cazaux, Thierry Lecroq, Eric Rivals. From Indexing Data Structures to de Bruijn Graphs. [Research Report] RR-14004, LIRMM. 2014. <lirmm-00950983v2>

Export

Share

Metrics

Record views

585

Document downloads

395