Translational Bioinformatics (Sandra Hellberg)

Question 1

RNA-sequencing analysis pipeline

Question 2

Raw count matrix

Question 3

Filtering and normalization of RNA-seq data

Answer 3

Filtering: We need to remove genes which have no expression in our data. There will be a lot of genes in the count matrix which have 0 genes. Removing genes with low counts decreases the problem with multiple testing. If we have 20,000 genes, you need to run 20,000 statistical tests and then a problem could be that you could get many false positives. Genes that are unexpressed in all samples have no biological meaning, so you remove the low count genes.

Question 4

Batch effects and biological confounders

Question 5

Batch correction and covariates

Question 6

Properties of RNA-seq data

Question 7

Differential expression analysis

Question 8

Multiple testing problem

Question 9

False positives (calculate fraction of false positives

Question 10

Multiple testing correction (FWER, Bonferroni, Benjamini Hochberg, FDR)

Question 11

Nominal p-values, adjusted p-values, q-values

Question 12

High-dimensonal data analysis (PCA, MDS, SVD, tSNE, K-Means, hierarchical clustering)

Answer 12

High dimensional data refers to a dataset in which the number of features is larger than the number of observations. The problem with this type of data is that it is very huge and quite computer heavy. Excel is not compatible with this. A lot of this sequencing data is stuff that you cannot understand.

Question 13

Biological pathways

Question 14

Gene enrichment analysis (pathway and gene set)

Question 15

Disease enrichment

Question 16

Different annotation data bases (KEGG, GO, DisGeNET, REACTOME)

Question 17

Measures of enrichment (effect size, fold enrichment, odds ratio)

Question 18

Fisher’s exact test