Introduction to RNA-seqWhat are we measuring in an RNA-seq experiment?Experimental design considerationsRNA-seq quantification: from reads to count matrixFinding the reference sequencesWhere are we heading towards in this workshop?
Figure 1

Figure 2

Figure 3

Figure 4

Figure 5
RStudio Project and Experimental Data
Figure 1

Figure 2

Figure 3

Figure 4

Importing and annotating quantified data into R
Figure 1
Exploratory analysis and quality control
Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Differential expression analysis
Figure 1

Figure 2

Figure 3
Shrinkage of log fold changes is useful for visualization and ranking of
genes, but for result exploration typically the
independentFiltering
argument is used to remove lowly
expressed genes.
Figure 4

Extra exploration of design matrices
Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Gene set enrichment analysis
Figure 1

Figure 2

Figure 3
Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12
