We are planning to compare expression patterns of long non-coding RNAs among different conditions in cell culture experiments. This will be performed by transcriptome sequencing on an Illumina HISeq platform. For normal mRNAseq we usually combine up to six samples with barcoded adaptors to save costs. Yet this decreases the sequencing depth per sample. Does anyone know how many samples could be combined on a HiSeq lane to get adequate results of lncRNA expression, taking the cost of sampling into consideration?
Rule of a thumb we apply: 100M reads per treatment group (ading all the replicates) to cover with reasonable depth a human (or vertebrate or complex plant) transcriptome.
So, accounting for some unmapped reads : 1 treatment group per lane (split in 3-4 biological replicates to make statistics happy) should be enough to see everything well. 2 Treatment groups per Illumina HiSeq lane (6-8 samples) may be enough for most count-based applications, but for cufflinks and lincRNA may not be enough. If you have paired reads: for lincRNAs it is better, for count-based expression checks worse (1pair=1reagment)
From a dataset of 400M reads we estimated that 95% coverage is achieved already at 100M reads and 97% at 200M reads . I agree with Michal that 1 lane per triplicates group is ok for most of RNAseq data on vertebrates with a PE 50bp. Perhaps on cell culture you can even multiplex more as you do not expect a big complexity.
Minimum of 10 MRs will get you plenty to play with. remember, there are 100s of annotated lncRNAs that few may be even working on; much less in your specific context!
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Biology
- Biomolecules
- Nucleic Acids
- RNA