[steven]

Thanks! Very useful paper for my purpose. It looks like if the run is RNA-Seq (or miRNA-seq etc...), the % coverage should be really low (the paper said that > 93% of uniquely mapped reads fell into exon regions). In this case, our numbers are funny . Oh well, dont know what is happening here.

It also depends on what kind of RNA you are working with (total RNA, polyA+ selected, cytoplasmic, nuclear, ribo-minus, etc). Plus this number of 93% is not a golden standard..
You may be interested in these previous threads: here and there

[dukevn]

See

http://www.genetics.wustl.edu/bio5488/wp-content/uploads/2010/01/Wold_NMeth.pdf

It's kinda old and only 25 bp reads but should give you a ballpark.

Thanks! Very useful paper for my purpose. It looks like if the run is RNA-Seq (or miRNA-seq etc...), the % coverage should be really low (the paper said that > 93% of uniquely mapped reads fell into exon regions). In this case, our numbers are funny . Oh well, dont know what is happening here.

[NextGenSeq]

See

http://www.genetics.wustl.edu/bio5488/wp-content/uploads/2010/01/Wold_NMeth.pdf

It's kinda old and only 25 bp reads but should give you a ballpark.

[dukevn]

In this case, i would say it is quite strange to get 92% of the genome covered by RNA-seq reads. Not?

Agreed. But I have no idea what number it should be. That is why I ask people if they do have some number for comparison.

[steven]

I meant percentage of mapped genome over the whole genome length, and it is mRNA-Seq.

In this case, i would say it is quite strange to get 92% of the genome covered by RNA-seq reads. Not?

[dukevn]

Are you aligning to the human or mouse genome or to only coding regions?

I mapped it to human (and mouse, we have both human and mouse samples) genome, and that is mRNA sample. Should I map it against the coding region only? How do I do that?

We run mostly paired end genomic DNA and we get 80 to 90% aligning to our reference genome (which contains coding and noncoding).

I guess your number is fine, since your is paired-end genomic DNA, right?

[NextGenSeq]

Are you aligning to the human or mouse genome or to only coding regions?

We run mostly paired end genomic DNA and we get 80 to 90% aligning to our reference genome (which contains coding and noncoding).

[dukevn]

The way I read it was 20 to 92% of his reads map to his reference genome depending on the lane. If its RNA-Seq hopefully those are mapping in the coding regions.

I meant percentage of mapped genome over the whole genome length, and it is mRNA-Seq. Does that mean lots of them are mapped in the coding regions, if the percentage of mapped genome is low (20-50%)?

As for the alignment score (percentage of reads mappable), we do have quite good number ranging from 80-90% for single end analysis (treated two ends like two single end runs), and about 60-70% for paired-end analysis.

Your numbers (80-90%) is paired-end alignment score or genome percentage?

Thanks,

D.

[NextGenSeq]

Only reads mapping on exon sequences. The RPMK value is normalized for total exon-length and the total number of matches in an experiment, in order to compare different experiments.

[thinkRNA]

" I got some numbers for genome coverage (percentage of mapping genome and whole genome length),"
ok, I thought he was talking about percent of genome covered.
Do you know when people calcluate RPKM, the million mapped reads is only from coding regions or the whole genome?

[NextGenSeq]

The way I read it was 20 to 92% of his reads map to his reference genome depending on the lane. If its RNA-Seq hopefully those are mapping in the coding regions.

[thinkRNA]

if only ~ 1-2 % of the genome is coding then isn't 92% genome coverage a bit too high? Am I missing something?

[NextGenSeq]

We usually get 25-30 million reads from a single lane of a flow cell. We usually run 2 x 72 bp. 80 to 90% map to the reference genome. The size of the human genome is 3.3 Gb and ~1 to 2% is coding.

If you get 92% of reads mapping thats pretty good, 20% not so good.

[dukevn]

It depends on the size of the genome, the number of lanes (or octets) and the length of the sequencing run (36 bp, 72bp, 100 bp etc)

Can you have some numbers for illustration? My dataset are Illumina 1 x 36, 2 x 36 with mouse and human samples, 7 lanes + PhiX. I really need some numbers to see how good or bad our runs are.

Thanks,

D.

[NextGenSeq]

It depends on the size of the genome, the number of lanes (or octets) and the length of the sequencing run (36 bp, 72bp, 100 bp etc)