viral genome alignment

[mathew]

I have a question, I am working on a chIPseq data where tumors are having a viral infection. We IP with a human specific antibody for our gene of interest. The S.E reads from hiseq were aligned to human genome using BWA which has worked fine and gave me some probable binding sites after peak calling. Now I am working on to find what happened to viral factors. So I took viral genome (around 10K) using Bowtie. Here is a screen shot for Bowtie SAM file, There are only 0.30% uniquely mapped reads.
QNAME FLAG RNAME POS MAPQ CIGAR MRNM MPOS ISIZE SEQ QUAL OPT
@HD VN:1.0 SO:unsorted
@SQ SN:AF148805 LN:137969
@PG ID:Bowtie VN:0.12.7 CL:"bowtie -q -p 8 -S -n 2 -e 70 -l 28 --maxbts 800 -y -k 1 -a --best --phred33-quals /tmp/3006527.cyberstar.psu.edu/tmp5nKzJC/tmpb50_zP /galaxy/main_pool/pool3/files/005/338/dataset_5338393.dat"
HWI-ST550_0201:3:1101:1671:2197#ACAGTG/1 4 * 0 0 * * 0 0 AAAATTCAGGCTCTCTATTTCACAGTTCATTAGTTCATTCGTTTACTGTG CCCFFFFFHHHHHJGIJJJJHIIJJJIGIHIIIJJGIJJJJJJJIJIJII XM:i:0
HWI-ST550_0201:3:1101:1678:2241#ACAGTG/1 4 * 0 0 * * 0 0 AGTGGTGTTTAATATAGTTTTGGGTATTTTTAACTAAAAATCATTGTTAT ?@@B?2AD?D<<CAE4AGHIF9CEG+AFDHID3C?9?CDFC**:?9*B9D XM:i:0
HWI-ST550_0201:3:1101:1626:2216#ACAGTG/1 4 * 0 0 * * 0 0 GTTGCGGGAGAAGCCAAACGCGGCGAGTCTTGCTAAAGCCGTCGCCGTAG BBCFFFFFFHHHF>GGGHCGEHIGGAE=CDFACEEEEDDDBDD;BB57<? XM:i:0
HWI-ST550_0201:3:1101:1580:2218#ACAGTG/1 4 * 0 0 * * 0 0 ACAGAAATGGCATCAAGAGACCTTGATTACAAGGATATGAATCTCTTAAG CCCFFFFFHHGHHIIJJIJJJJJJJDIJJJIIIJIJJJJIJJIJIJJIJI XM:i:0
HWI-ST550_0201:3:1101:1779:2214#ACAGTG/1 4 * 0 0 * * 0 0 CCAATCTCTGCTACAGTTTGTTTCCCTCAATTTCTAATTACTTTAAAAAG CC@FFFFFHHDHDFGHEGIJIIJJJJGIGJJJJJIIJJEIIEHGJIGJJI XM:i:0

_________________________________________________________

How I should be selecting only uniquely mapped reads to viral genome?
Why I have so low number of uniquely mapped reads? Is there any way that I can increase this unique mapping? What will be the best strategy to align to viral genome in this case, Should I be aligning to viral genome all reads or first align to human then align to un-mapped reads to viral genome. I also tried it with BWA gave around 0.29% unique alignment.

[gsgs]

I don't know your programs and acronyms, but what I woulf do is
checking subsequences of length -say- 12. How many % of them are
found, this is easy to check even for big databases. (all viruses
from genbank or such)

[mathew]

Consider me a new in viral genome. Could you please explian how can I calculate sub seq ratio? Any pointer to URL or guidence will be great. I am using Bowtie aligner.

Thanks

[gsgs]

well, I'm the new one since I don't know about the normal software and databases and companies etc.
I write my own software, mainly just for influenza.

After some years (!) I noticed, that for most comparisons we don't need
alignment, we can just count the number of matching subsequences of certain length, no matter at what position they appear.

I think this is also basically used in "blast", why it's so fast for big databases.

So I wrote a program for that, (Windows 32-bit,cmd.exe commandline - DOS)
but presumably there are other programs available for UNIX,Win64, etc,

I can send my program, with source code or I run your data through it
(all genbank viruses) it finds matching subsequences length 15-28

(I speculate this is what you want, but am not sure)

[mathew]

gsg Thanks I have sent a private message with my email. If you can share your script that will be a good start for me.

[swbarnes2]

Ideally, if you know your samples has both human and viral sequence in it, you should be aligning to a reference that has both. That will give you the most accurate alignment.

Would you expect the antibody to be binding to viral sequence?

If you are getting very few reads aligning to virus, the simplest explanation is that you have very little virus in your library. Why have you dismissed that possibility?

[mathew]

Thanks swbarnes2,

I agree, so what I have done is I have aligned with just viral genome that is not very good (around 0.3% aligned reads). Then when I aligned to a combined human and viral genome it gives me around 70% aligned reads. However I am not sure how I can separate out the reads aligned to viral genome only. It may be that most of reads are mapped to human. As far as no of viral particles in that library are concerned the experiment has worked and mostly the induction is by viral infection which has translated well into human genome analysis. Any thought on how we can I get viral genome reads only? I have read papers which state that they just aligned to viral genome and then called peaks but it is not working in my case unless I am missing something.

Thanks

[Gators]

There are bowtie parameters to output only uniquely aligning reads.

I think it is -m 1, the only output alignments are unique

[swbarnes2]

Originally Posted by mathew

Thanks swbarnes2,

I agree, so what I have done is I have aligned with just viral genome that is not very good (around 0.3% aligned reads). Then when I aligned to a combined human and viral genome it gives me around 70% aligned reads. However I am not sure how I can separate out the reads aligned to viral genome only.

Samtools view is how you do that.