What are common errors in RNA-Seq analysis that skew results?

Question

[crri]

Good morning,

I'm looking at a set of RNA-Seq data that was generated by a previous postdoc and my process was fairly simple:

1. Convert BCL to FASTQ.GZ files with bcl2fastq
2. BioJupies-based automated pipeline

While I'm new to this field, these results don't seem to make biological sense. For a sanity check, I downloaded some raw RNA-Seq data from a paper and performed the same pipeline, which produced results in-line with the results of the paper. I know it isn't much to work from, but based on this, does anyone happen to know what could be going wrong?

Also, I used this pipeline because it was easy and I didn't want to introduce any errors from my self-taught ignorance. I've seen suggestions of a STAR -> featureCounts -> DESeq2 method, but I'm not sure if that is the current best available approach. What is the current gold-standard for RNA-Seq analysis? Are there any tutorials/guides you felt were helpful when you started learning this?

Best,
CRRI

EDIT:

The experiment was a pilot to see if our approach could detect differences in gene expression of bone marrow-derived macrophages exposed to treatments of either LPS treatment or co-culture with LL/2 cells. We had a time-course of a week. A clear example was when we compared 7 day exposure vs sham treatment in wildtype cells. The top 3 gene ontology results were for downregulated ribosome biogenesis, rRNA processing, and rRNA metabolic process and upregulated neutrophil degranulation, neutrophil activation, and neutrophil mediated immunity. The top 3 KEGG pathways were for downregulated ribosome biogenesis, endocytosis, and biosynthesis of amino acids and upregulated apoptosis, proteoglycans in cancer, and VEGF signalling. While this may be flawed, to the best of my understanding, treatment with LPS should activate canonical inflammatory processes. I'd be glad to add any additional details if it would help.

[rpolicastro]

What didn't make sense about the results?

[crri]

As an example for the bone marrow derived macrophagess treated with LPS (control vs 4 day treatment), the top 3 gene ontology results were for upregulated ribosome biogenesis, rRNA processing, and rRNA metabolic process and downregulated neutrophil degranulation, neutrophil activation, and neutrophil mediated immunity (among others). Given the LPS treatment, wouldn't inflammatory processes be upregulated?

Answer 1

[swbarnes2]

Come on. No one can answer anything about a dataset whose problems we can't see that went through a pipeline we know nothing about.

STAR-> FeatureCounts is fine, though I'd prefer RSEM instead of FeatureCounts, because it's smarter about dealing with ambiguous reads.

I think Kallisto -> DESeq might be a little more popular now than genome aligners like STAR, but that doesn't mean that STAR is wrong to use.

[crri]

Understandable. I'll add a bit to the original post to make things clearer. And, thanks for your description of pipelines!