Transcriptomics: A short introduction to the core concepts of microarrays and RNA sequencing

A short introduction to the core concepts of transcriptomics, which is the systematic measurement of all transcripts. I will start with quickly explaining the relation between genes (DNA) and transcripts (RNAs), then cover the two main technologies (DNA microarrays and RNA sequencing) including important steps in the computational analysis, important differences in experimental sample preprocessing (poly(A) selection vs. rRNA depletion), and finally statistical analysis to identify differentially expressed genes or transcripts.
0:00 Introduction: gene transcription, definition of transcriptomics, and overview of the presentation
1:09 Microarrays: cDNA / two-channel arrays, competitive hybridization, Affymetrix GeneChips / one-channel arrays, and non-linear normalization
3:05 RNA sequencing: read mapping (HISAT2 / STAR), transcript assembly (StringTie), read count matrix (HTSeq), and normalized metrics (RPKM)
4:32 Sample preprocessing: most RNA is ribosomal, polyA selection of transcripts, rRNA depletion, ncRNAs, and total RNA-seq
5:58 Statistical testing: differential expression, statistics for microarrays (normal distribution), and statistics for RNA-Seq data (negative binomial distribution)

[@aewe4239]

thanks for the great talk. I may ask if you can do some lectures on node/edge rewiring across multiple networks. Let's say I have 3 different groups and I have 3 different networks. What nodes/edges are rewiring across these conditions etc?

[@larsjuhljensen]

Great suggestion! The main problem I see with rewiring analysis is, that it mainly makes sense to do, if you have actual condition-specific networks. By that I mean networks in which the edges themselves relate to conditions so that actual rewiring takes place. Unfortunately, most of the networks that people refer to as "condition-specific networks" in fact only have conditional data for the nodes. You are thus looking at a networks, for example from STRING, which are the same global network filtered down purely based on the expression of the nodes. For such datasets, I am not convinced that rewiring analysis makes much sense, since you are in fact looking at nodes coming and going (and taking edges with them), so it is debatable if the network as such is being rewired.

[@aewe4239]

@@larsjuhljensen Thank you so much for your reply.