Please help: imperfect reference genome/get consensus on genome/read alignment?

KAP · 08-18-2011, 01:17 PM

I an doing RNA-seq on two closely related plant species (~98% similar in transcript region), one of which is sequenced.

My problem comes when I align the reads from the unsequenced species to the sequenced genome. Most of the reads align (at 2 bp mismatch level), however some are being rejected due to 3 or 4 bp mismatches. I would like to improve on the alignability of reads/reference genome to ensure the observed differences in expression between species are true (not due to differences in read alignability).

My idea is as follows:

Align the reads to the reference (allow ~2 bp mismatch)
Use the alignment to generate a consensus sequence, ie where there are SNPs between the reference and the reads, change the base on the chromosome to reflect that of the reads. Use stringent parameters so only true SNPs are 'corrected'.
Realign the reads to the modified genome (allow ~2 bp mismatch). Now reads with 2 bp mismatches should align 'perfectly' and reads with up to 4 bp mismatches will align.

Can anyone recommend programs or a strategy to do this, or even comment on the validity of the approach?

Any feedback/contribution would be much appreciated! Thank you!!

Note: the reads are very short: 40 bp.

macrowave · 08-19-2011, 08:14 AM

I had the similar idea when facing a somewhat similar situation. In theory, this would work, but in practice, I found the transcript levels quantified from the 'reconstructed' genome are biased towards lower end. I suspected that these didn't reflect the true expression levels in the 'reconstructed' species, skewed towards underestimation. I haven't thought about the exact causes, but I think it's related to the 'reconstruction' process and the heterogeneity of genomic divergence along chromosomes between two species. So in the end, I just mapped to the same reference, but allowing extra mismatches for the divergent species. Another way to look into this is just going ahead and to do a de novo assembly for the divergent species, but that opens a can of worms, too. And it's a complete different ball game.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
hg19 genome reference for short read mapping	yh253	Bioinformatics	4	12-29-2013 10:11 PM
Alignment to selected region of the reference genome	houkto	General	1	02-20-2012 06:51 AM
Assisted de novo genome assembly? Create new consensus mapping reads to reference?	zmartine	Bioinformatics	8	02-10-2012 01:31 AM
alignment and reference genome assembling	slny	Bioinformatics	3	03-16-2011 07:49 AM
Reference genome for MAQ - split reference genome by chromosome or not?	inesdesantiago	Bioinformatics	4	02-18-2009 09:44 AM

08-18-2011, 01:17 PM	#1
KAP Member Location: Adelaide, Australia Join Date: Mar 2010 Posts: 12	Please help: imperfect reference genome/get consensus on genome/read alignment? I an doing RNA-seq on two closely related plant species (~98% similar in transcript region), one of which is sequenced. My problem comes when I align the reads from the unsequenced species to the sequenced genome. Most of the reads align (at 2 bp mismatch level), however some are being rejected due to 3 or 4 bp mismatches. I would like to improve on the alignability of reads/reference genome to ensure the observed differences in expression between species are true (not due to differences in read alignability). My idea is as follows: Align the reads to the reference (allow ~2 bp mismatch) Use the alignment to generate a consensus sequence, ie where there are SNPs between the reference and the reads, change the base on the chromosome to reflect that of the reads. Use stringent parameters so only true SNPs are 'corrected'. Realign the reads to the modified genome (allow ~2 bp mismatch). Now reads with 2 bp mismatches should align 'perfectly' and reads with up to 4 bp mismatches will align. Can anyone recommend programs or a strategy to do this, or even comment on the validity of the approach? Any feedback/contribution would be much appreciated! Thank you!! Note: the reads are very short: 40 bp.

08-19-2011, 08:14 AM	#2
macrowave Member Location: New York Join Date: May 2010 Posts: 13	I had the similar idea when facing a somewhat similar situation. In theory, this would work, but in practice, I found the transcript levels quantified from the 'reconstructed' genome are biased towards lower end. I suspected that these didn't reflect the true expression levels in the 'reconstructed' species, skewed towards underestimation. I haven't thought about the exact causes, but I think it's related to the 'reconstruction' process and the heterogeneity of genomic divergence along chromosomes between two species. So in the end, I just mapped to the same reference, but allowing extra mismatches for the divergent species. Another way to look into this is just going ahead and to do a de novo assembly for the divergent species, but that opens a can of worms, too. And it's a complete different ball game. Last edited by macrowave; 08-22-2011 at 07:51 AM.