How quantitative is contig coverage?

[AlexB]

Hi All,
I am working on draft (resequencing) genomes of about 2Mb. In one run we multiplexed three strains and that works quite ok (as far as i can see now).
I noticed one thing which I find odd and which might be an artefact or meaningful observation.
When I plotted the sequence coverage (contrib to consensus) I noticed that in my draft which has on average about 20-25 fold coverage that some contigs have extreme high coverage (over that complete contig). this can be as high as 125 fold coverage. .... Aaah repeats..... no.
I looked at it and one example for instance is of size 1kb containing one complete ORF. But there are more regions similar to this.

Does this mean this ORF is about 4-5 fold present in the genome or is this a too simple conclusion. Could there be amplification or assembly artefacts???

Thanks for sharing any thoughts on this!

Alex

[kmcarr]

Alex,

What is the ORF? This is obviously a bacterium (2 Mbp genome) and there are duplicated genes observed in bacteria, most notably the rRNA genes. It could also be extrachromosomal (i.e. plasmid) DNA.

[AlexB]

coverage quantitative?

Hi kmcarr,
yes indeed its a bacterial (streptococcus) genome. I guess that since for transcriptomics the technique is quantitative so my guess it should be for genomes as well. Just wondered whether such "artefacts" might be the result of library generation and or the used chemistry for sequence analysis (I am not so much in those details).
Regarding the 4-5 fold more coverage I indeed expect the example to be a multiple duplicated gene (have to look for transposon elements and so) or indeed as you say extra chromosomal like plasmid. The example (I looked it up was a transposase of 1163bp and flanking 270 post and 100 pre). Another contains IS630-spnl. Most indeed transposable elements.... But how to find where they belong????? Looking at the assembly and flanking sequences/SNPs???

In addition I just finished some pre-processing of another strain this afternoon that also has an average coverage of roughly 20-35 times but has several larger regions that have coverage over multiple contigs in synteny with about 100 fold. More extreme even a region with even 800 coverage (really smells like plasmid...?). Don't know yet what they contain but I hope to know soon.

So is it generally true that the coverage ratio approx resembles copy number per genome???

Thanks
ALex

[flxlex]

Your extra-high coverage contig should indeed represent a repeated piece of genome, as the 454 assembly program (newbler) collapses repeats. To find out where it belongs, you can have newbler generate the contig graph file (using the -g option on the command line). This file contains information on the nodes, i.e. the contigs (length, coverage) and on the edges, i.e. the (number of) reads that cross the boundary between contigs (start in one contig and end in another). For your high-coverage contig, you will find that it has several neighbors on both ends. Your next task is to figure out which neighbors belong together (with the high-coverage contig in between), by for instance PCR. There might be sequence variants among the different copies of the repeated region, and sequencing the PCR products will tell youwhich variant belongs where. Of course, this works best for contigs that are within the length that can be sequenced by Sanger reads...

Or, you might want to invest in paired-end sequencing to place the repeats... Apparently, one run from an 8kb jumping paired end library will get you one scaffold (ordered, oriented contigs with estimates of gap sizes) for bacterial genomes (we are about to test this for one of our strains):

http://www.454.com/downloads/DeNovoSmallGenomes_Flyer.pdf

Good luck!

[AlexB]

quant coverage

Thanks flxlex,
good suggestions. I indeed intended tolok at the assembly and coverage of that specific contigs. I just find the numbers for repeats so high. In the latter 800 fold coverage was a transposase IS4.....
Didn't know newbler could do that. I will install the software shortly on our linux machine and see how it goes.
Alex