Skip to content

The Key idea

In Reference-based RNAseq analysis, the read counts are treated as proxies to RNA steady state levels

This idea implies 3 steps

1. Map reads to a reference genome with aligners such as:

  • TopHat2
  • HiSat2
  • STAR

→ These aligners are “splice aware”

→ They generate a BAM Alignment file

2. Use a software tool and genome annotations to count the reads spanning genes or transcripts

  • Input files for the counting tools are mostly the BAM Alignment file generated at step 1.
  • Mostly used counting tools:
    • FeatureCounts
    • HTseqCounts
    • Some aligners (STAR, Salmon, ...) are embedding their own counting procedure

  • Genome annotations

    They are an essential element of the whole analysis: Keep in mind that you only count what you told the counting tool to count!

    Most often genome annotations are provided in a GTF format. GFF3 or BED12 may also be used by some counting tools.

    Given the importance of annotations in Reference-based Expression analysis it is highly recommended that you take some time to well understand the structure of a GTF file, its links with the genome version as well as its own version. In particular annotation versions evolve more rapidly that the genome version to which they are linked. There are several years between different releases of genome assembly, whereas different versions of the genome annotation may be separated by only several months (curation, new gene discovery, etc.)

3. Use a tools to detect statistically significant changes of read counts between conditions

Most used statistics tools (R packages):

  • DESeq2
  • EdgeR