Masked/Unmasked Reference Genome

ytmnd85 · 04-13-2009, 12:56 PM

Hey everyone,
I just started working with Illumina sequencing, and while preparing to have my sample's reads mapped to a reference genome, I was asked whether I would like my reads mapped to a 'masked' reference genome or an 'unmasked' reference genome.

Could someone explain to me the fundamental difference in masked and unmasked reference genomes, and which would be better to use? The goal of our sequencing is to compare transcription levels between two samples.

swbarnes2 · 04-13-2009, 01:54 PM

"Masked" in general means that repetative parts of your reference genome are hidden away (turned into n's), so they won't be aligned to. Repeats, especially with short read technology, can be hard to deal with, since you can't unambiguously place a read whose sequence aligns to multiple places in the genome (though paired ends reads can help, if one end is anchored in unique sequence)

With transcription, repeats aren't as big a problem, since most of the really repetative stuff in a mammalian genome isn't in the transcriptome, though there are still going areas where two different genes share identical sequence, and there's just not much you can do about that.

If you are trying to count transcript levels, you probably want to stick to counting the reads that only fall in unique sequence, so I'd think that masking is a good idea.

Are you using the whole genome as the reference, or a putative transcriptome? Reads that span between two exons won't align to the whole genome, so you might undercount.

ytmnd85 · 04-13-2009, 03:32 PM

I've had Illumina align the reads to the whole genome. I'm working with soybean, so I wasn't sure if a thorough transcriptome had been published.

In regards to the reads that will span two exons and therefore will not align, will this undercounting produce a significant problem, or will there still be, in general, plenty of reads that will fall within the one exon?

BioWizard · 04-14-2009, 09:59 AM

I would not recommend masking the reference, as it is just one more step that loses information (as if we don't have enough already, with sequencing errors, and imperfect references). Especially since the "masking" cannot be 100% perfect. You can cause your alignment to think some sequences are unique whereas they are not. If you want to do some filtering - better do it AFTER alignment. Jusy my humble opinion.

cgb · 05-31-2009, 09:45 AM

i agree with biowizard. smaller search space does not guarantee higher accuracy/sensitivity.

Annette · 05-31-2009, 04:52 PM

masking the reference genome can in the worst case rather lead to an increased incidence of falsely mapped reads:
reads that would have a best hit in a masked region will tend to be mapped to some other place in the genome where the next best (but still incorrect) hit is found.
This of course depends on your mapping stringency – you won´t probably have the problem if you only allow for perfect matches.

On the other hand you run into a problem measuring expression levels when keeping repetitive regions.
Some software packages place ambiguous reads randomly which makes it impossible to measure transcript abundance for the respective genes.

It can be of help if after mapping you can still trace which reads were ambiguously mapped. Then you can apply filtering (after alignment).

But this is a problem already for duplicated genes, not just for low complexity regions...

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Targeted Genome Assembly for region poorly represented in reference genome?	gumbos	Bioinformatics	1	01-09-2012 05:01 PM
Please help: imperfect reference genome/get consensus on genome/read alignment?	KAP	Bioinformatics	1	08-19-2011 08:14 AM
transferring annotations from reference genome to the resequenced genome	mike.t	Bioinformatics	1	09-17-2010 06:35 AM
Masked or unmasked genome for ChIP-seq analysis?	hbbio	Bioinformatics	3	04-07-2009 12:14 PM
Reference genome for MAQ - split reference genome by chromosome or not?	inesdesantiago	Bioinformatics	4	02-18-2009 09:44 AM

04-13-2009, 12:56 PM	#1
ytmnd85 Member Location: Nebraska Join Date: Apr 2009 Posts: 10	Masked/Unmasked Reference Genome Hey everyone, I just started working with Illumina sequencing, and while preparing to have my sample's reads mapped to a reference genome, I was asked whether I would like my reads mapped to a 'masked' reference genome or an 'unmasked' reference genome. Could someone explain to me the fundamental difference in masked and unmasked reference genomes, and which would be better to use? The goal of our sequencing is to compare transcription levels between two samples.

04-14-2009, 09:59 AM	#4
BioWizard Member Location: Houston, TX Join Date: Mar 2009 Posts: 27	masking reference I would not recommend masking the reference, as it is just one more step that loses information (as if we don't have enough already, with sequencing errors, and imperfect references). Especially since the "masking" cannot be 100% perfect. You can cause your alignment to think some sequences are unique whereas they are not. If you want to do some filtering - better do it AFTER alignment. Jusy my humble opinion.

04-13-2009, 01:54 PM	#2
swbarnes2 Senior Member Location: San Diego Join Date: May 2008 Posts: 912	"Masked" in general means that repetative parts of your reference genome are hidden away (turned into n's), so they won't be aligned to. Repeats, especially with short read technology, can be hard to deal with, since you can't unambiguously place a read whose sequence aligns to multiple places in the genome (though paired ends reads can help, if one end is anchored in unique sequence) With transcription, repeats aren't as big a problem, since most of the really repetative stuff in a mammalian genome isn't in the transcriptome, though there are still going areas where two different genes share identical sequence, and there's just not much you can do about that. If you are trying to count transcript levels, you probably want to stick to counting the reads that only fall in unique sequence, so I'd think that masking is a good idea. Are you using the whole genome as the reference, or a putative transcriptome? Reads that span between two exons won't align to the whole genome, so you might undercount.

04-13-2009, 03:32 PM	#3
ytmnd85 Member Location: Nebraska Join Date: Apr 2009 Posts: 10	I've had Illumina align the reads to the whole genome. I'm working with soybean, so I wasn't sure if a thorough transcriptome had been published. In regards to the reads that will span two exons and therefore will not align, will this undercounting produce a significant problem, or will there still be, in general, plenty of reads that will fall within the one exon?

05-31-2009, 09:45 AM	#5
cgb Member Location: Cambridge Join Date: May 2008 Posts: 50	i agree with biowizard. smaller search space does not guarantee higher accuracy/sensitivity.

05-31-2009, 04:52 PM	#6
Annette Junior Member Location: Europe Join Date: May 2009 Posts: 3	masking the reference genome can in the worst case rather lead to an increased incidence of falsely mapped reads: reads that would have a best hit in a masked region will tend to be mapped to some other place in the genome where the next best (but still incorrect) hit is found. This of course depends on your mapping stringency – you won´t probably have the problem if you only allow for perfect matches. On the other hand you run into a problem measuring expression levels when keeping repetitive regions. Some software packages place ambiguous reads randomly which makes it impossible to measure transcript abundance for the respective genes. It can be of help if after mapping you can still trace which reads were ambiguously mapped. Then you can apply filtering (after alignment). But this is a problem already for duplicated genes, not just for low complexity regions...