low featurecounts assignment rate despite of good mapping with STAR

Question

[Eunbi]

Hello all, I checked the questions posted on the forum, but couldn't find the answer I was looking for, so I'm leaving a question.

I'm doing total RNA sequencing and have a question about my results. There are three types of data: from AML12, 3T3L1 and mouse testis. (All data is mouse)

I did mapping with genome using STAR, and counting with featureCounts.

The ratio of uniquely mapped reads of all data is between 80~90%, but only 3T3L1 featurecounts result seems different.

AML12 and testis transcriptome showed assigned ratio between 70~80%, but 3T3L1 transcriptome showed only 58~62% and majority is "No features".

All data were analyzed using the same gencode.vM30.annotation gtf file and GRCm39 genome sequence, and the code used is below.

STAR:

./STAR --runThreadN 16 --genomeDir /path/to/genomeindex/dir --alignendsType EndToEnd --readFilesIn sample-1.fastq sample-2.fastq --outSAMtype BAM SortedByCoordinate --outFileNamePrefix sample

featureCounts:

./featureCounts -T 16 -M --fraction -p -a /path/to/gtf -o /path/to/output /path/to/input/*.bam

Does anyone know why only 3T3L1 has a low counting ratio?

Answer 1

[colindaven]

I would check the ribosomal RNA fraction using https://github.com/hzi-bifo/RiboDetector

Also, try to find the coverage peaks in your (where have the reads been mapped to) using eg sambamba coverage or one of many others. You can sort the file to find the highest coverage regions.

Also, just eyeball your fastq for the poor sample - does it look repetitive ? Blast a few suspicious seqs.

[Eunbi]

Thank you for your quick reply. I think I need to check rRNA fraction first, because I didn't. Also, I'll look into the coverage peaks as you suggested. Thank you very much.