low assignment reads % in RNA-seq

Question

[parinv]

I performed RNA-seq analysis on 8 samples(2 control and 6 samples) . I used featureCounts function (Rsubread package) to count the number of reads. I found that Number of assigned reads were very poor. I am attaching the table here for your reference. I used following function:

fc<- featureCounts(annot$Sample,isPairedEnd=TRUE, GTF.attrType = "transcript_id",GTF.featureType="exon", annot.ext = "GCF_000203855.3_ASM20385v3_genomic.gtf", nthreads = 3, isGTFAnnotationFile = TRUE)

I do not understand why assignment% is low, when alignment % is found higher. Can anyone help me think of possible reasons for why I'm getting such low values for assignment?

Thanks a lot. Parin

[cpad0112]

check if multi-mapped reads are counted or not and use correct strandedness. Talk to the core and check if ribodepletion worked for all the samples.

[lakhujanivijay]

can you post the alignment summary metrics for example like this

[parinv]

I figured out that using GTF.featureType="exon" is giving low assignment %, If I use GTF.featureType="CDS", the assignment % is increasing as shown here. I cannot decide which feature type should I use. I read some answers 1 2 and found that any one can be used. Kindly share some information regarding use of exon or CDS.

Thankyou

[lakhujanivijay]

This is subjective - pertaining to questions asked in the experiment

Still confused about exons versus CDS

Answer 1

[solomoncharles77]

One possible reason is that you may need to indicate the strand information of your library. Besides, this issue has been discussed in multiple threads, like here and here. Those threads and others should contain suggestions that can help you resolve the issue. Good luck.

Answer 2

[sysboolean]

Can you post more information about the origin of the samples ? Someone I know once performed RNA-seq on what they thought was human cell culture, but got poor alignment percents like yours. Further investigation (BLASTing a few reads against nr db) showed that the cells were from a non-human mammal. Might help to run Fastqscreen to make sure your samples were not switched during sequencing.