I’m trying to count reads mapped to human tRNA in some 1x40bp illumina small-seq files previously published. Since normal genome reference files do not include tRNAs, this means I need to make my own for analysis. The reference tRNA fasta file I used to create my genome index I took from gtrnadb using high-confidence tRNAs identified in hg38 assembly (hg38-tRNAs.fa) and mapped to the genome annotation file from ENCODE genes (Predicted tRNA genes GTF/GFF3 files). I trimmed using fastp and got back decent enough reads to continue. I ran bowtie2 and got back roughly 6% uniquely matching, which I reasoned may be more related to biology than a technical issue.
The problem: When I feed this bam file and relevant reference files into htseq-count, however, I get back “__no_feature” and “__not_aligned” reads.
The things I’ve tried:
- Try other aligners and mappers, like STAR and hisat and featureCounts
- List the “feature” flag as “tRNA” to reflect the GTF file column 3
- Altered strandedness for each possibility
- Confirmed chromosomal tags are consistent (however, this doesn’t necessarily relate because according to my tRNA reference fa file, each tRNA is being read as its own chromosome. I confirmed this with Samtools: IdxStats)
- Viewed the bam mapping in IGV and confirmed reads are associated with specific tRNAs
- Removed the top 5 # lines in the GTF file and altered the attribute to “GTF” as recommended in other help requests
- Using normalizeFasta to “standardize” my reference fasta genome by wrapping to 80 and removing anything after the first whitespace. This truncates every line to, for example, “>Homo_sapiens_tRNA-Ala-AGC-1-1” and not the more typical “chrX” in other reference genomes
- Tried setting my custom genome as a custom build, and set my cleaned fasta file and GTF to that build. featureCounts no longer has any unmapped reads, but still shows only “unassigned_NoFeatures” reads
- Downloaded alternative tRNA fasta/GTF files from genome.ucsc.edu tables which fixed the “feature” issue being tRNA instead to exon, as expected in default layouts
I still think this is a chr mapping issue or an issue due to substantial multimapping? But I’ve exhausted all of my google skills to getting this far. Would really appreciate any help.
Thank you for your time!