Need advice for pre process and assembly

[patouch74]

Hi,

I have two paired end files and I'm trying to merge them in order to assemble them, so I have two strategies but I don't what to use and I'm not really sure if they are correct or not.

1) First method

- Join them with fastq-join
The outputs are 3 files: one refers to the joined reads from the original files (JOIN), one is the unjoined reads alone from the 1 (PE_1), and one likewise from 2 (PE_2).
- Assembly with velveth :
velveth Dir 31 -shortPaired -separate -fastq PE_1.fq PE_2.fq -short -fastq JOIN.fq
- velvetg etc.


2) Second method
- Join them with velvet-shuffleSequences_fastq.pl
The output is : one file containing both reads in interleaved format (OUT)
- Assembly with velveth :
velveth Dir 31 -shortPaired -fastq OUT.fq
- velvetg etc.

Some people had done these before ?
Thanks

[Brian Bushnell]

If you want to merge them, I suggest using BBMerge, which has a very low false positive join rate. False joins cause assembly errors.

bbmerge.sh in1=PE_1.fq in2=PE_2.fq out=merged.fq outu=unmerged.fq

The "outu" file will contain unmerged reads interleaved. But, I encourage you to try assembling twice, once with the original reads and once with the merged + unmerged reads, because merging is not guaranteed to improve assembly; sometimes it will make it worse.

And as for "velvet-shuffleSequences_fastq.pl", not sure what the point is of that. Interleaving paired reads won't affect your assembly.

[patouch74]

Thanks for your answer,
I will try your soft. To assembly merged + unmerged reads, the correct command is :
velveth Dir 31 -shortPaired -fastq merged.fq -short -fastq unmerged.fq

right ?

and for velvet-shuffleSequences_fastq.pl, I've read

http://wiki.bioinformatics.ucdavis.edu/index.php/Illumina_assembly_using_velvet

[Brian Bushnell]

I'm guessing that information is obsolete, as Velvet can handle paired reads in two files just fine now.

But the command would be:

velveth Dir 31 -shortPaired -fastq unmerged.fq -short -fastq merged.fq

...since "unmerged" contains paired reads while "merged" contains the single reads. Also, you will probably get a better assembly with a higher K than 31.

[patouch74]

Brian, do you suggest to trim bad quality sequences before join or after join ?

[Brian Bushnell]

I suggest this order:

1) removal of phiX, and other artifact/contaminant reads
2) adapter trimming
3) normalization and/or error-correction and/or subsampling (all optional, depends on whether you have too much coverage)
4) merging (optional)
5) quality-trimming of unmerged reads only, to around Q10
6) assembly

All of this, aside from assembly, can be done with BBTools. Steps 1, 2, and 5 can be done with bbduk.sh and step 3 can be done with bbnorm.sh (error-correction or normalization) or reformat.sh (subsampling).

1)
bbduk.sh -Xmx1g in1=read1.fq in2=read2.fq out=clean.fq ref=phix174_ill.ref.fa.gz hdist=1 k=31

2)
bbduk.sh -Xmx1g in=clean.fq out=trimmed.fq ref=truseq.fq.gz ktrim=r mink=11 hdist=1 k=25

3) (optional)
ecc.sh -Xmx29g in=trimmed.fq out=corrected.fq

4) (optional)
bbmerge.sh in=corrected.fq out=merged.fq outu=unmerged.fq

5)
bbduk.sh -Xmx1g in=unmerged.fq out=qtrimmed.fq qtrim=rl trimq=10 minlength=50

6)
velveth Dir 31 -shortPaired -fastq qtrimmed.fq -short -fastq merged.fq

The reference files phix174_ill.ref.fa.gz and truseq.fq.gz are both included with BBTools, in the "resources" directory. If you do error-correction (which will improve the rate of merging), the "-Xmx29g" flag is just an example; rather than 29g, it should be set to around 85% of the computer's physical memory.