DBG2OLC:Efficient Assembly of Large Genomes Using Long Erroneous Reads of the Third Generation Sequencing Technologies
=======
Our work is published in Scientific Reports:
Ye, C. et al. DBG2OLC: Efficient Assembly of Large Genomes Using Long Erroneous Reads of the Third Generation Sequencing Technologies. Sci. Rep. 6, 31900; doi: 10.1038/srep31900 (2016).
http://www.nature.com/articles/srep31900
The manual can be downloaded from:
https://github.com/yechengxi/DBG2OLC/raw/master/Manual.docx
To use precompiled versions,please go to:
https://github.com/yechengxi/DBG2OLC/tree/master/compiled
To compile, download the source code of each program into separate folders and use:
g++ -O3 -o DBG2OLC *.cpp
https://github.com/yechengxi/SparseAssembler
g++ -O3 -o SparseAssebmler *.cpp
https://github.com/yechengxi/Sparc
g++ -O3 -o Sparc *.cpp
It is important to follow the steps before you invent your own pipeline. As combining DBG2OLC with other software tools may produce worse results. The reason is that many existing assembly/error correction tools introduce errors.
We have provided code to help you select a subset of the reads:
https://github.com/yechengxi/AssemblyUtility
After compilation, you can use the following command to select a subset of reads from fasta/fastq files. Note that longest 0 is used here, if you set it to 1 it will select the longest reads.
./SelectLongestReads sum 600000000 longest 0 o Illumina_50x.fastq f Illumina_500bp_2x300_R1.fastq
./SelectLongestReads sum 260000000 longest 0 o Pacbio_20x.fasta f Pacbio.fasta
And you can use the following command to evaluate an assembly.
./AssemblyStatistics contigs YourAssembly.fasta
The program will generate two txt files containing essential statistics about your assembly.
Please make sure they are the raw DBG contigs without using repeat resolving techniques such as gap closing or scaffolding. Otherwise you may have poor final results due to the errors introduced by the heuristics used in short read assembly pipelines.
SparseAssembler command format:
./SparseAssembler GS GENOME_SIZE NodeCovTh FALSE_KMER_THRESHOLD EdgeCovTh FALSE_EDGE_THRESHOLD k KMER_SIZE g SKIP_SIZE f YOUR_FASTA_OR_FASTQ_FILE1 f YOUR_FASTA_OR_FASTQ_FILE2 f YOUR_FASTA_OR_FASTQ_FILE3_ETC
A complete example on the S.cer w303 dataset:
Download the Illumina reads from
ftp://qb.cshl.edu/schatz/ectools/w303/Illumina_500bp_2x300_R1.fastq.gz
or here:
http://pan.baidu.com/s/1sk9bEnN
Normally with ~50x coverage, NodeCovTh 1 EdgeCovTh 0 can produce nice results.
./SparseAssembler LD 0 k 51 g 15 NodeCovTh 1 EdgeCovTh 0 GS 12000000 f ../Illumina_data/Illumina_50x.fastq
In this test run, the N50 is 29 kbp. As we have selected the beginning part of the sequencing file, which can be of lower quality, the next step may help to improve the assembly quality.
For other more complex genomes or a different coverage, the first run may not generate the best result. The previous computations can be loaded and two parameters can be fine-tuned to construct a cleaner de Bruijn/ k-mer graph:
./SparseAssembler LD 1 NodeCovTh 2 EdgeCovTh 1 k 51 g 15 GS 12000000 f ../Illumina_data/Illumina_50x.fastq
The N50 is improved to 32kbp in my run.
The output Contigs.txt will be used by DBG2OLC.
Feed DBG2OLC with the contig file in fasta format from the previous step (Contigs.txt in this example).
Download the PacBio reads from:
ftp://qb.cshl.edu/schatz/ectools/w303/Pacbio.fasta.gz
or here:
http://pan.baidu.com/s/1sk9bEnN
The basic command format of DBG2OLC is:
./DBG2OLC k KmerSize AdaptiveTh THRESH_VALUE1 KmerCovTh THRESH_VALUE2 MinOverlap THRESH_VALUE3 Contigs NGS_CONTIG_FILE f LONG_READS.FASTA RemoveChimera 1
In the following example, the first 20x PacBio reads are extracted from the abovementioned file and we can assemble with:
./DBG2OLC k 17 AdaptiveTh 0.0001 KmerCovTh 2 MinOverlap 20 RemoveChimera 1 Contigs Contigs.txt f ../Pacbio_data/Pacbio_20x.fasta
In our test run, the N50 is 583kbp.
There are three major parameters that affect the assembly quality:
M = matched k-mers between a contig and a long read.
AdaptiveTh: adaptive k-mer matching threshold. If M < AdaptiveTh* Contig_Length, this contig cannot be used as an anchor to the long read.
KmerCovTh: fixed k-mer matching threshold. If M < KmerCovTh, this contig cannot be used as an anchor to the long read.
MinOverlap: minimum overlap score between a pair of long reads. For each pair of long reads, an overlap score is calculated by aligning the compressed reads and score with the matching k-mers.
At this point, the parameters may be fine-tuned to get better performance. As with SparseAssembler, LD 1 can be used to load the compressed reads/anchored reads.
Suggested tuning range is provided here:
For 10x/20x PacBio data: KmerCovTh 2-5, MinOverlap 10-30, AdaptiveTh 0.001~0.01.
For 50x-100x PacBio data: KmerCovTh 2-10, MinOverlap 50-150, AdaptiveTh 0.01-0.02.
Some other less flexible or less important parameters:
k: k-mer size, 17 works well.
Contigs: the fasta contigs file from existing assembly.
MinLen: minimum read length.
RemoveChimera: remove chimeric reads in the dataset, suggest 1 if you have >10x coverage.
For high coverage data (100x), there are two other parameters:
ChimeraTh: default: 1, set to 2 if coverage is ~100x.
ContigTh: default: 1, set to 2 if coverage is ~100x.
These two are used in multiple alignment to remove problematic reads and false contig anchors. When we have high coverage, some more stringent conditions shall be applied as with the suggested parameters.
Install blasr and the consensus module (sparc/pbdagcon). Make sure they are in your path variable. The input files for consensus are:
(1) backbone_raw.fasta by DBG2OLC
(2) DBG2OLC_Consensus_info.txt by DBG2OLC
(3) DBG contigs (in fasta format)
(4) PacBio reads (in fasta format)
You can check the N50 of (1) to see if you are satisfied, otherwise keep tuning and don¡¯t proceed.
// this is to concatenate the contigs and the raw reads for consensus
cat Contigs.txt pb_reads.fasta > ctg_pb.fasta
// we need to open a lot of files to distribute the above file into lots of smaller files
ulimit -n unlimited
//run the consensus scripts
sh ./split_and_run_sparc.sh backbone_raw.fasta DBG2OLC_Consensus_info.txt ctg_pb.fasta ./consensus_dir 2 >cns_log.txt
The A. thaliana Ler-0 dataset:
20x PacBio reads:
./DBG2OLC KmerCovTh 2 AdaptiveTh 0.005 MinOverlap 20 RemoveChimera 1 Contigs Contigs.txt k 17 f ../PacBio/20x.fasta
40x PacBio reads:
./DBG2OLC KmerCovTh 2 AdaptiveTh 0.01 MinOverlap 20 RemoveChimera 1 Contigs Contigs.txt k 17 f ../PacBio/40x.fasta
The H. sapiens dataset:
Longest 30x PacBio reads:
./DBG2OLC k 17 KmerCovTh 2 MinOverlap 20 AdaptiveTh 0.01 RemoveChimera 1 Contigs Contigs.txt f 30x.fasta >DBG2OLC_LOG.txt
The program command is slightly different for purly Illumina reads assembly.
Example command: ./DBG2OLC LD 0 MinOverlap 70 PathCovTh 3 Contigs Contigs.txt k 31 KmerCovTh 0 f ReadsFile1.fa f ReadsFile2.fq f MoreFiles.xxx
There are four critical parameters:
k: k-mer length (max size: 31).
KmerCovTh: explained above, suggest 0-1.
MinOverlap: explained before.
PathCovTh: the minimum occurrence for a compressed read for a compressed read to be used, suggest 1-3.
Assembly is reported as DBG2OLC_Consensus.fasta.
The command we used for E. coli Illumina Miseq dataset:
./DBG2OLC k 31 PathCovTh 2 MinLen 50 MinOverlap 31 Contigs Contigs.txt KmerCovTh 0 f reads.fasta
If you compile DBG2OLC in your environment and it gets stuck when running, try the precompiled version we included.
There are many reads correction techniques available for 3GS assembly. Not all of them are doing the right job. Correction may lead to worse results as structural errors may be introduced. So if you decide to use a correction algorithm, please compare results with/without correction.
Most short read assemblers introduce structural errors to reach higher contiguity. These errors can lead to poor final results. DBG2OLC requires accurate short read assembly to start with. Other than SparseAssembler, a few reported successful assemblers are: Platanus, Meraculous.
Since version 5.1, blasr need double dash "--" for some options : "In release v5.1 of BLASR, command-line options will use the single dash/double dash convention: Character options are preceded by a single dash. (Example: -v) Word options are preceded by a double dash. (Example: --verbose)"
So if you are using the latest version of blasr, the blasr calling command in split_and_run_sparc.sh (or other similar scripts) should be modified to:
cmd+="blasr --nproc 64 ${split_dir}/${chunk}.reads.fasta ${split_dir}/${chunk}.fasta --bestn 1 -m 5 --minMatch 19 --out ${split_dir}/${chunk}.mapped.m5;
DBG2OLC is a single threaded program. The most time consuming step is to calculate the compressed reads. If you are trying to assemble a large genome, below is an easy way to speed up:
(1) Split the 3GS reads into several batches.
(2) Run DBG2OLC with the same set of parameters and Illumina Contigs in each of these directories.
In dir1:
./DBG2OLC k 17 AdaptiveTh 0.005 KmerCovTh 2 MinOverlap 20 RemoveChimera 1 Contigs ../Contigs.txt f ../Pacbio_data/Pacbio_batch1.fasta
In dir2:
./DBG2OLC k 17 AdaptiveTh 0.005 KmerCovTh 2 MinOverlap 20 RemoveChimera 1 Contigs ../Contigs.txt f ../Pacbio_data/Pacbio_batch2.fasta
...
(3) Move all the compressed reads (filenames starting with "ReadsInfoFrom_") into one folder (say dir3). If you happen to put the raw PacBio reads into separate folders, put them also in dir3 or make sure they can be located in the command when you make the final call in step 4.
(4) Rerun DBG2OLC with such modifications: "LD 1" to and include all the compressed reads.
In dir3:
./DBG2OLC k 17 LD 1 AdaptiveTh 0.01 KmerCovTh 2 MinOverlap 30 RemoveChimera 1 Contigs ../Contigs.txt f ../Pacbio_data/Pacbio_batch1.fasta f ../Pacbio_data/Pacbio_batch2.fasta