Feedback on Fastqc of trimmed quantseq data

[laurahspencer]

Looking for some feedback on my expression data post-filtering. Data was generated using the QuantSeq 3' mRNA kit, so libraries were generated near the 3' end of transcripts, and sequencing was on a NovaSeq, single-read, 100bp.

Here's the MultiQC of trimmed reads. I used cutadapt to trim universal Illumina adapter, poly-A and poly-G tails, and quality-filtered for min. quality = 15, and min length 20 bp. Here's my Jupyter notebook where I trim/filter, align, and generate counts/gene.

Things I need to address prior to confidently assessing diff. expression:

• Why the weird GC content?
• Duplicates - should I try to remove using Picard? Reads are all supposed to be clustered near the 3' end, so there's a possibility I could over-duplicate.
• Should I remove all genes ID'd in my No-Tissue-Control (sample Register interest in using UniProt annotation system in standalone manner #571) from all samples?
• Should I remove the low-count samples prior to DESeq (44, 35)?

[laurahspencer]

Individual FastQC files here: https://github.com/laurahspencer/O.lurida_QuantSeq-2020/tree/master/qc-processing/fastqc/trimmed