Awesome
RNAseq_analysis_scripts
Scripts for RNA seq analysis. They can be used directly using the outputs from [IARCbioinfo's RNA seq workflow].(https://github.com/IARCbioinfo/RNAseq-nf)
Unsupervised analysis: RNAseq_unsupervised.R
This script performs unsupervised analyses (Principal Component Analysis and clustering) from htseq-count outputs.
Prerequisites
This R script requires the following packages:
- ConsensusClusterPlus
- ade4
- DESeq2
- fpc
Usage
Rscript RNAseq_unsupervised.R [options]
| PARAMETER | DEFAULT | DESCRIPTION |
|---|---|---|
| -f | . | folder with count files |
| -o | out | output directory name |
| -p | count.txt | pattern for count file names |
| -n | 500 | number of genes to use for clustering |
| -t | auto | count transformation method; 'rld', 'vst', or 'auto' |
| -c | hc | clustering algorithm to be passed to ConsensusClusterPlus |
| -l | complete | method for hierarchical clustering to be passed to ConsensusClusterPlus |
| -h | Show help message and exit |
For example, one can type
Rscript RNAseq_unsupervised.R -f input -p count -t rld -o output/ -n 500
Details
The script involves 3 steps
- Data transformation using either variance-stabilization log2-tranform or the r-log tranform from package DESeq2
- PCA of tranformed counts with package ade4
- Clustering of transformed counts with package ConsensusClusterPlus
Output
- a .RData file with 4 objects: transformed counts di, the ade4 pca object pca,the ConsensusClusterPlus object clusters, and the list of cluster and item consensus icl
PCA
- PCA plots with the first 2 PC, with colors corresponding to the clusters from ConsensusClusterPlus, for K=1,2,...,6 clusters.
- loadings plot for the first 2 PC
- csv file with the genes with the greatest loadings in the first PC
Clustering
- a folder with plots from ConsensusClusterPlus
Compare an unsupervised analysis with a list of variables: RNAseq_unsupervised_compare.R
This script compares the result of an unsupervised analyses (Principal Component Analysis and clustering) obtained for example using script RNAseq_unsupervised_compare.R with an arbitrary number of variables (categorical or continuous).
Prerequisites
This R script requires the following packages:
- ade4
- permute
Usage
Rscript RNAseq_unsupervised_compare.R [options]
| PARAMETER | DEFAULT | DESCRIPTION |
|---|---|---|
| -R | . | .RData file with results from clustering in variable clusters and results from PCA in variable pca |
| -i | . | name of input file with variables in column and variable names as first line |
| -m | 2 | minimum number of clusters |
| -M | 5 | maximum number of clusters |
| -o | out | output file preffix |
| -h | Show help message and exit |
For example, one can type
Rscript RNAseq_unsupervised_compare.R -R RNAseq_unsupervise.RData -i variables.txt -o output/
Details
For each clustering present in variable cluster, the script involves 3 steps
- Plotting the variables on the first two PC axes
- Computing the best matching between clusters and the levels of the variable (note: if the variable is continuous, categories are formed by subdividing the range of the variable into intervals of the same size)
- Testing the significance of the best matching by computing a null distribution of matchings across 1000 permutations
Output
For each column (i.e., variable) of the input table, a .pdf file with K rows, where K is the number of clusterings in variable cluster, and 3 columns:
- Column 1 represents the first 2 PCs of the PCA; colored ellipses correspond to the clusters of the clusters variable from the .RData file, and point colors correspond to the levels of the variable
- Column 2 represents the best matching between the clustering and the variable
- Column 3 represents the null distribution of the best matching (gray), the observed best matching and the p-value of the best matching (red)
Supervised analysis: RNAseq_supervised.R
This script performs supervised analyses (Differential Expression Analysis) at the gene level from htseq-count outputs.
Prerequisites
This R script requires the following package:
- DESeq2
- gtools
Depending on the options used, the following packages are also required:
- BiocParallel (with option -c)
- IHW (with option -m)
Usage
Rscript RNAseq_supervised.R [options]
| PARAMETER | DEFAULT | DESCRIPTION |
|---|---|---|
| -f | . | folder with count files |
| -g | . | file with sample groups |
| -o | out | output directory name |
| -p | count.txt | pattern for count file names |
| -c | 1 | number of cores for statistical computation |
| -q | 0.1 | False Discovery Rate |
| -m | FALSE | Use Independent Hypothesis Weighting for multiple-testing procedure |
| -h | Show help message and exit |
For example, one can type
Rscript RNAseq_supervised.R -f input -g groups.txt -o output/
Details
The script performs DE analysis of gene count data under a Poisson glm with package DESeq2. When multiple groups are present (e.g., A, B, and C), computes results for contrasts corresponding to all combinations of 2 groups (A vs B, A vs C, and B vs C).
Output
- a plot of fold changes as a function of normalized counts
- plots of normalized counts of the most significant DE gene, as a function of the group, for each pair of group levels
- .csv files with gene names, fold changes, and p-values, for each combination of 2 groups
- a .RData file with 1 object: the deseq2 results table
Supervised analysis: RNAseq_supervised_transcript.R
This script performs supervised analyses (Differential Expression Analysis) at the transcript level from StringTie outputs.
Prerequisites
This R script requires the following package:
- ballgown
- genefilter
- dplyr
Usage
Rscript RNAseq_supervised_transcript.R [options]
| PARAMETER | DEFAULT | DESCRIPTION |
|---|---|---|
| -f | . | folder with folders of sample input files |
| -g | . | file with sample groups |
| -o | out | output directory name |
| -p | . | pattern for input folder names |
| -t | 1 | Threshold variance in gene expression (FPKM) |
| -r | Row names for group file | |
| -c | 2 | Column index of covariable to use for regression; other columns are treated as adjustment variables |
| -h | Show help message and exit |
For example, one can type
Rscript RNAseq_supervised_transcript.R -f input -g groups.txt -o output/
Details
The script performs DE analysis of transcript FPKM data under a hierarchical linear model with package ballgown, optionally correcting for additional variables.
Output
- plots of expression levels for each gene with DE transcripts
- a .csv file with gene names, IDs, p-values, and q-values
- a .RData file with 1 object: the ballgown object