question about polishing in wtpoa-cns

[ruanjue]

question about polishing in wtpoa-cns

[ruanjue]

For short reads, wtpoa-cns can success in multiple alignment, but the consensus codes may not cope with short reads. Thanks for your question, I think I will find time to make it works on short reads SAM from bwa or other alignment tools.

Jue

[TinaH10]

I just ran it using my Illumina data to polish and it ran all the way through. It is hard to judge what it has done, as there is no stats file or such, that summarizes the changes after polishing. Or is there one? I will align some of the contigs and see what the differences are with polishing with Illumina. I'll update what I find.

[ruanjue]

Sorry for the last email, I made a mistake.

[ruanjue]

Please update 7aa6f87

• View the multiple alignments

wtpoa-cns -t 1 -d dbg.ctg.lay.fa -i <(samtools view dbg.ctg.lay.fa.srt.bam) -w 200 -j 150 -rS 2 -R 0 -N 40 -vv 2>&1 | less -S

• Run consensus

wtpoa-cns -t 32 -d dbg.ctg.lay.fa -i <(samtools view dbg.ctg.lay.fa.srt.bam) -w 200 -j 150 -rS 2 -R 0 -N 40 -fo dbg.ctg.lay.polished.fa

Please note that current code doesn't pay effort on heterozygote, you will find it make wrong decision on heter sites. So, please give me more time to find a good consensus calling method to perform well on both long noisy reads and short accurate reads.

Jue

[ruanjue]

New consensus mode: dp-call_cns works on short reads polishing now.

wtdbg-2.3/wtpoa-cns -t 64 -d dbg.ctg.lay.fa -i <(samtools view dbg.ctg.lay.fa.srt.bam) -fo dbg.ctg.lay.fa.samlay.fa -b 1 -c 1 -w 200 -j 150 -N 40 -R 0 -rS 2

From my test in yeast, quast5 reported improvements on base accuracy.

# mismatches per 100 kbp     161.80        52.48
# indels per 100 kbp         129.97        23.67

[TinaH10]

Excellent, thank you!

[TinaH10]

Oh, one more question: Will it also correct assembled contigs based on paired end information? For example, if the Illumina PE reads consistently map to different contigs, will the contig be broken up accordingly? Or will it just correct errors in sequence?

[ruanjue]

That is interesting, but better to leave it to other tools.