deBGA is a seed-and-extension-based read alignment tool. It is suitable for aligning various kinds of high-throughput sequencing reads to multiple similar reference genomes.
deBGA indexes the genome through de Bruijn graph framework, which captures the genomic substrings implied by the unipaths and indexes them with a hash table-based index. With the index, deBGA adopts a seed-and-extension strategy. In seeding phase, a series of seeds from the reads are matched to the reference genome inferring a set of putative read positions (PRPs) from the matched positions. In the extension phase, the reads are aligned with the local sequences around the PRPs to compose full read alignments.
deBGA has outstanding throughput on aligning reads from various prokaryote and eukaryote genomes. A series of benchmarks on model organism genomes, e.g., E. coli, H. sapiens, etc., demonstrated that it can simultaneously achieve good throughput, sensitivity and accuracy in various kinds of read alignment tasks. deBGA is open source and free for non-commercial use.
deBGA is mainly designed by Bo Liu and developed by Hongzhe Guo in Center for Bioinformatics, Harbin Institute of Technology, China.
The memory usage of deBGA can fit the configurations of most modern servers and workstations. Its peak memory footprint depends on the length of reference genome, i.e., 40.32 Gigabytes and 9.07 Gigabytes respectively for the real H. Sapiens and E. Coli genomes, on a server with Intel Xeon CPU at 2.00 GHz, 1 Terabytes RAM running Linux Ubuntu 14.04.
The wall time and memory footprint of the index construction for the references (the k-mer size of the index is 22) is as follows. The time is in seconds, the memory footprints are in Gigabytes, and ref.file size is in Gigabytes.
No. Reference Time Memory ref.file size
1 E. coli K-12 MG1655 substrain 21 2.02 0.004
2 10 E. coli strains 116 3.02 0.05
3 20 E. coli strains 200 3.02 0.09
4 40E. coli strains 280 4.03 0.19
5 62E. coli strains 729 5.04 0.29
6 A. thaliana reference genome (TAIR10) 972 7.05 0.11
7 19 A. thaliana strains 4993 11.09 2.12
8 RefSeq bacteria genomes 69912 153.30 8.56
9 xenograft model (hg19+MM10) 42851 130.03 5.53
10 human reference genome (hg19) 15786 69.55 2.94The memory footprint of deBGA when aligning the reads from the listed datasets (the alignment is conducted with the default setting of deBGA) is as follows. The memory footprint are in Gigabytes.
No. Dataset Memory
1 E. coli Sim-62 9.07
2 ERR008613 9.07
3 SRR522163 9.07
4 SRR530851 9.07
5 A. thaliana Sim-19 12.1
6 Meta Sim 80.64
7 Xenograft Sim 60.48
8 Meta Pseudo-real 80.64
9 H. sapiens Sim-100 40.32
10 H. sapiens Sim-250 40.32
11 ERR174324 40.32
12 Hiseq X Ten: NA12878_L3 40.32
13 ERR161544 40.32Current version of deBGA needs to be run on Linux operating system.
The source code is written in C, and can be directly download from: https://github.com/HongzheGuo/deBGA
The makefile is attached. Use the make command for generating the executable file.
deBGA index [options] reference.fasta \<index_route>
Index reference in RdBG-Index format
deBGA aln [options] \<index_route> \<single_end_read.fastq [pair_end_read1.fastq pair_end_read2.fastq]> \<result_file.sam>
Align read to its primitive location in Reference
deBGA index
-k, INT the k-mer length of the vertices of RdBG. This is a basic parameter for building theRdBG-index. For the current version of deBGA,
the range of -k parameter is restricted to 21-28 bp, considering both of the effectiveness of the seeds and memory footprint[22].
deBGA aln
-k, INT the minimum length of a valid Uni-MEM seed. For current version of deBGA, this setting should be equal to the k-mer length of the RdBG-index[22].
-s, INT the number of iterations of re-seeding. deBGA iteratively aligns a read in at most (-s + 1) iterations with various set of seeds.
This parameter works combining with the minimum interval of seeding (the -i option) and the maximum allowed number of hits per seed (the -n option).
That is, in the r-th iteration (r = 1 ,…, -s), deBGA tries to generate seeds at every ((-s)– r +1)*(-i) bp along the read.
If the read still cannot be successfully aligned after -s iterations, deBGA would ignore -n option to handle very repetitive reads in the (-s+1)-th iteration[4].
-i, INT the minimum interval of seeding. This parameter determines the density of seeds, which is related to the sensitivity and efficiency of alignment.
Configuring this parameter with lower value will make deBGA generate seed more densely, which could improve the sensitivity, but at the expense of throughput[5].
-n, INT the maximum allowed number of hits per seed. In the first -s iterations of the alignment process, the seeds with more than -n hits would be discarded for achieving faster speed.
DeBGA ignores this restriction to introduce repetitive seeds if the read still cannot be successfully aligned after -s iterations[300].
-c, NUM the threshold on the edit distance for early stop. In each iteration, deBGA checks the edit distance of the obtained best alignment.
If the ratio ED_best/RL <(-c), where ED_best and RL are respectively the edit distance of the best alignment and the read length,
deBGA considers that the read is confidently aligned and early-stops the alignment[0.05].
--cl, NUM the adjusted threshold on the edit distance for early stop. When --cl option is set, in any given iteration, if there is at least one Uni-MEM seed available for extension,
but no successful alignment is obtained, the threshold on the edit distance(-c) can be dynamically adjusted to the value of --cl in next iterations.
This is a heuristic may acceleratethe alignment ofdivergent reads, e.g., reads having many low quality bases.
If --cl is not set, there will be no change on the -c option during the process (default: following the setting of -c).
--local, the local alignment option for confident alignment. When --local option is set, in any given iteration, if there is at least one Uni-MEM seed available for extension,
but no successful alignment is obtained,deBGA perform local alignment instead of end-to-end alignment in following iterations.The best obtained local alignment will be output as the result.
It is also worthnoting that the --cl option and --local option should not be simultaneously set (default: not set).
--local-match, INT the score for a matched base in the local alignment.This option will take effect only if --local option is set[1].
--local-mismatch, INT the penalty for a mismatched base in the local alignment. This option will take effect only if --local option is set[4].
--local-gap-open, INT the penalty for a gap open in the local alignment.This option will take effect only if --local option is set[6].
--local-gap-extension, INT the penalty for gap extension in the local alignment.This option will take effect only if --local option is set[1].
--stdout, (default: not set) output alignments by stdout. This option will let deBGA directly output alignments by stdout instead of user defined file.
-u, INT the upper limit of insert size. For a pair-end read, deBGA pairs the alignments of the two ends according to the upper (-u option) and lower (-f option) limits of the insert size.
deBGA will consider it as a suitable pair-end alignment only if the inferred insert size is within the range [-f, -u][1000].
-f, INT the lower limit of insert size. For a pair-end read, deBGA pairs the alignments of the two ends according to the upper (-u option) and lower (-f option) limits of the insert size.
deBGA will consider it as a suitable pair-end alignment only if the inferred insert size is within the range [-f, -u][50].
-o, INT the maximum number of alignment output.deBGA outputs at most -o alignments for the read. This is except for the pair-end reads which are handled with the anchoring alignment strategy.
For thosereads, the number of outputsis determined by the -x option[20].
-x, INT the maximum number of alignment output for anchoring alignment. For the pair-end reads aligned with the anchoring alignment strategy, deBGA will output at most -x alignments[150].
-l, INT the maximum allowed read length. For the current version of deBGA, reads shorter than -l bp will be normally processed, andfor reads longer than -l bp, only the first -l bp will be aligned,
and the other parts will be trimmed. Set -l option with a larger number may slightly increasethememory footprint. For most nowadays next generation sequencing reads, e.g., reads from Illumina platforms,
the default setting is long enough to work withoutthe trimming. Moreover, the current version ofdeBGA can support reads upto 4096 bp (setting -l to 4096)[512].
-e, INT the budget for single-end alignment. In single-end read alignment, deBGA sets a budget on the computation resource in advance for balancing the efficiency and the sensitivity. More precisely, in the extension phase,
deBGA subsequently extend the candidate seeds in order of their coverage lengths, until more than -e extension operations have been totally executed after handling some of the seeds, or all the seeds are extended[100].
-p, INT the number of threads. The current version of deBGA supports upto 32 threads in read alignment[1]. Genome indexing: deBGA index Reference Index_Dir
Read alignment: deBGA aln Index_Dir Fastq_File Sam_file
We simulated a series of datasets from various genomes, i.e., human genome build GRCh37/hg19 and the reference sequences of the 62 E. Coli strains, through Mason Simulator (version0.1.2). For human genome, five read lengths from 100 bp to 250 bp were used and the mean and standard deviation of the insert size are respectively 500 bp and 25 bp for all the datasets. For the E.coli strains, 100 bp Illumina-like pair-end reads were simulated for evaluation. These datasets helped us to evaluate the performance of deBGA. The datasets have been uploaded to Google Drive, and can be downloaded through the following link:
deBGA: read alignment with de Bruijn Graph-based seed and extension.
For advising, bug reporting and requiring help, please contact ydwang@hit.edu.cn; hzguo@hit.edu.cn