Help!!!!!! RNA-Seq

[guzhi100]

Hi all,

We did RNA-Seq using RNA isolated from six E11.5 mouse embryonic hearts. And we put all six samples in one lane for Illunima Hiseq 2000 RNA-Seq. Each sample contains around 800 ng total RNA. rRNA was depleted before RNA sequencing.

As far as I know, one lane can produce around 150 million reads. My aim is to see differentially expressed isoforms.

I am worrying if the depth coverage for each of my samples is enough.
Did I make a wrong decision to put all six samples in one lane? The reason for this decision is that we thought we already deleted rRNA which normally eat up most of the reads. The mRNA left in the sample won't need so much reads. This is why we put all six samples in one lane.

Could you please let me know your opinion? Do we still have enough depth coverage for isoforms?

Thanks

[scbaker]

At 25M per sample, my guess is that you're on the low side of the number of reads you want. It also depends on how you 'depleted' your sample of rRNA. The most 'aggressive' way of doing this is with some sort of poly dT pulldown to isolate the polyA fraction. With that method, 25M would be ok, but you'll likely only see the medium to highly expressed isoforms. If you used RiboMinus, you may find yourself very short of the ideal # of reads, both because some of your reads will be dedicated to ncRNAs and also because RiboMinus tends not to deplete all of the rRNA. (I've heard that Ribo-Zero is better, but I don't know how much better). My guess is that you might want 2-3X more reads per sample (or more if you want the rare isoforms). Curious to hear what others think.

[chadn737]

Hi all,

We did RNA-Seq using RNA isolated from six E11.5 mouse embryonic hearts. And we put all six samples in one lane for Illunima Hiseq 2000 RNA-Seq. Each sample contains around 800 ng total RNA. rRNA was depleted before RNA sequencing.

As far as I know, one lane can produce around 150 million reads. My aim is to see differentially expressed isoforms.

I am worrying if the depth coverage for each of my samples is enough.
Did I make a wrong decision to put all six samples in one lane? The reason for this decision is that we thought we already deleted rRNA which normally eat up most of the reads. The mRNA left in the sample won't need so much reads. This is why we put all six samples in one lane.

Could you please let me know your opinion? Do we still have enough depth coverage for isoforms?

Thanks

Multiplexing all 6 samples is the right approach. It is the optimal design to run all your replicates in the same lane to minimize effects across lanes.

~25 million reads per sample, you will be able to detect differential expression in your more highly expressed genes. But for more lowly expressed genes, your depth will likely be too low.

However, since you multiplexed your samples (the right thing to do), if you find you do not have not sequenced enough, then the solution is simple, just run all six samples on another lane to get the extra depth you need.

[guzhi100]

Hi, thanks so much for your answers, which are really helpful.

[billstevens]

I have almost the exact same concern. Thanks very much, the comments have been very useful. I analyzed my 3 samples results using cuffdiff, and what I was wondering was, what is the minimum number of mapped fragments for a gene to still make statements about its expression?

[billstevens]

bump?

Sorry to be pest, if this isn't the most interesting question, but I really would love some more info on it. I had a run with my 3 samples on one lane of HiSeq1000. I'm about to run my duplicates, and my lab can only afford one more lane. What I'm wondering is, should I repeat the first run and put three samples in one lane? Or, should I put 6 samples (2 sets of the 3 conditions) on one lane?

I'm trying to balance the loss of coverage depth versus duplicates. I'm using human epithelial cells.

To make this determination, I thought I would need to look at my data. I have about 400 "significantly" differentially expressed genes. I have about 100 M mapped hits per condition. So for a FPKM of 1, I guess I would have 100 counts for that gene. If I ran twice as many samples, I guess my FPKM would be 0.5, so with a total count of 50, could I still make determination on gene expression?

[pbluescript]

If all you are looking to do is get differentially expressed genes, you are better off sequencing more samples to help estimate biological variability.

[billstevens]

Thanks very much!