Aligning to HIV reference

[hifer]

Hi everyone,

First time poster here

I'm trying to align some Illumina paired reads (88bp) to the HIV1 reference (NC_001802.1, ~10k bp) using Bowtie2. However, somehow I found that these reads only align to the first half of the reference (refer to attachment).

This seems to be consistent across aligners (Bowtie2 & BWA), so I'm suspecting that it might be a characteristic of HIV. Has anyone come across any similar findings or suggestions?

Thanks!

[swigut]

[...]I'm suspecting that it might be a characteristic of HIV. Has anyone come across any similar findings or suggestions?

Thanks!

Unlikely, the only ambiguous part of the sequence are the LTRs.
1. Check your sequencing result manually if you have reads from the other half
2. Check your alignment statistics, do most of the reads align?
3. Check if the genome sequence you used to build the index was complete.

Tomek

[Andrei D.]

What kind of sample was used for library preparation? Maybe it was amplicon?
(then only part of HIV genome was sequenced)

[cswarth]

What parameters are you passing to bowtie when doing the alignment?
I'm curious because I am doing something similar, aligning deep sequencing reads from a v3 portion of env to NC_001802.

If I use the default bowtie2 parameters (--end-to-end mode), I get no alignments at all. However if I align in --local mode, the middle of my reads align but about a quarter of the read is soft-clipped at each end.

bowtie2 -x ../references/NC_001802 --local -N 1 -f -U sequences.fasta -S alignment.sam

I'm trying to figure out if this means I did not properly clean up the adapters on each end, or if there is just too much variation from the reference that bowtie is never going to be able to do a good job.

[cswarth]

To follow-up on this notion of using bowtie2 to align sequences to an HIV-1 reference -

My experience is that aligning HIV-1 sequences using Bowtie 2 does not work well. I suppose it depends on the reference and on the region being sequenced, but the diversity of HIV genomes, on top of sequencing errors, is often too great to be aligned by seed-and-extend techniques used by bowtie. A colleague suggested using smith-waterman alignment, or profile HMMs.

Below is an image taken from IGV showing the diversity of the gp-120 sequences I am working with. (Each color bar represents a position where the alignment differs from the reference.) These sequences were aligned to NC_001802 using a codon-aware smith-waterman aligner . It is this diversity, and divergence from the reference, that make it so hard for bowtie2 to do the alignment.

[hifer]

Hi everyone,

Sorry for the hiatus after posting this. Just as an update, I've managed to figure out why my reads are only aligning to the first half of HIV1, it's due to the way the library was prepared.

Apparently my collaborators have only sequenced the gag pol region of the virus (as this region is highly conserved) and therefore the reads only align to this particular region of the viral chromosome, no surprises afterall!

Thanks everyone for their input!