RNA seq PE vs SE read length

[Tom_C]

Hello everyone,

We are about to send some samples off for sequencing, and I would like some input before we spend all of our money.

We have RNA from a wild type and a mutant bacteria, with four biological replicates each (eight samples total). We wish to examine all the differentially expresses genes between the genotypes, and I was wondering what is the best read length to use?

Our genome is sequenced and assembled (4.9 Mbp), and we will be generating rRNA-depleted stranded libraries to be multiplexed on a single Illumina HiSeq lane.

What would be the optimal method to sequence? 50bp vs 100bp, SE vs PE?

[Brian Bushnell]

50bp SE would be sufficient for the vast majority of genes in most bacteria. 2x100bp is always better, of course, but only marginally for bacterial RNA expression.

But if you want, you can use KmerCountExact like this:

kmercountexact.sh in=genome.fa k=50 khist=stdout

For E.coli, I got this output:

#Depth  Count
1       4536984
2       11541
3       5674
4       1023
5       1692
6       1635
7       4075
8       30
9       117
10      961
11      68

For 4639675 bases, 4536984 kmers occurred exactly once in the genome. That means 97.8% of error-free 50bp reads will map uniquely. It only increases to 4552752 for k=100, which is not much of an improvement. Longer, paired reads are more useful when dealing with differentially-spliced or highly repetitive eukaryotes.