Overrepresentation of Reads Mapping to Intronic Regions

[Matthieu-Duot]

Hello,

I'm working with paired-end data from different projects (different origins). I wanted to reanalyze these data using my own pipeline with STAR (version 2.7.10a). I was able to obtain correct mapping, with an average of 80% uniquely mapped reads for most of my samples. However, when I started working on these data, I noticed that in the ReadsPerGene.out.tab file, a majority of my reads (~2/3) were classified as N_noFeature. Using rnaseqc, I observed that these reads were mapped to intronic regions.

Here are some relevant statistics:

Exonic Rate: 0.358933
Intronic Rate: 0.478841
Intergenic Rate: 0.0854653
Intragenic Rate: 0.837774

Since the intergenic rate seems reasonable, I don't think there was DNA contamination in the samples. I have the impression that the issue might be related to an annotation problem. I used a genome index generated from Ensembl files, and I tried generating another index using Gencode files, but the result was the same.

Here is the STAR command I used:

Singularity> STAR --runThreadN 8 \
--genomeDir Genome_Dir/Genome_index \
--readFilesIn Test_GenDir/subset_R1_231024.fastq.gz Test_GenDir/subset_R2_231024.fastq.gz \
--readFilesCommand zcat \
--outFilterType BySJout \
--outFilterMultimapNmax 20 \
--alignSJoverhangMin 8 \
--alignSJDBoverhangMin 1 \
--outFilterMismatchNmax 999 \
--alignIntronMin 20 \
--alignIntronMax 1000000 \
--alignMatesGapMax 1000000 \
--sjdbGTFfile Genome_Dir/gtf/Homo_sapiens.GRCh38.112.gtf \
--outReadsUnmapped None \
--outFileNamePrefix Test_newSTAR/InitialGenDir/InitGD \
--outSAMtype BAM SortedByCoordinate \
--outSAMunmapped None \
--outBAMsortingThreadN 4 \
--quantMode GeneCounts \
--outSAMattributes NH HI NM MD AS nM jM jI XS

Do you have any idea why this might be happening?

Thanks for your help.