TiSA: TimeSeriesAnalysis - A transcriptomic analysis tool for both RNA sequencing and microarray data

TimeSeriesAnalysis (TiSA) is an analysis and visualization package for RNAseq and microarray data. TS extracts significant genes from time course transcriptomic data by performing differential gene expression on both the conditional and temporal axes. It then employs partitioning algorithm based on recursive thresholding (PART) clustering to identify small genomic clusters of relevance, followed by running the clusters through gprofiler to reveal the biological relevance of each cluster.

• Data normalization and processing
• PCA plots
• Differential gene expression (conditional and temporal)
• PART clustering
• Heatmaps for both differential expression summary and clustering results
• Trajectory of identified clusters
• Gprofiler (functional enrichment) analysis of clusters
• Dotplots and MDS plots of Gprofiler results
• Nearest ancestor clustering of GOs
• GO ancestor queries

You can install the stable version of TimeSeriesAnalysis like so:

install.packages("devtools")
devtools::install_github("Ylefol/TimeSeriesAnalysis@master")

The development version can be installed by replacing ‘master’ with ‘dev’.

Devtools is required for the installation of this package. To install Devtools, visit the Devtools main page

Certain bioconductor packages will have to be installed before installation of TimeSeriesAnalysis. The code snippet below gives the method along with some of the packages that should be installed via this method.

if (!require("BiocManager", quietly = TRUE))
  install.packages("BiocManager")
bio_pkgs <- c('DESeq2','GOSemSim','GO.db','limma','ComplexHeatmap',
'AnnotationDbi','GenomeInfoDb','GenomicRanges','Biobase','S4Vectors',
'BioCGenerics','MatrixGenerics','IRanges','BiocFileCache'
'SummarizedExperiment','org.Hs.eg.db','org.Mm.eg.db','org.Ce.eg.db')
BiocManager::install(bio_pkgs)

TimeSeriesAnalysis was developed with user-friendliness in mind, the core script of the package is a Rmarkdown file with explanations of what each code block performs. Computationally intensive code blocks save their work in order to avoid loss of computation time. TimeSeriesAnalysis comes with a test dataset. Users are recommended to first run the package with the test data to then modify the Rmarkdown script with the necessary information for their purposes (input files, organism, genes of interest etc…). The rmarkdown script and the expected result from an example run can be found and downloaded from this repository, they are located in the ‘rmarkdown_method’ folder. If users prefer a script approach, the equivalent of the rmarkdown report is provided in two scripts, one for the computation tasks and a second for the analyses. These two scripts are found in the ‘script_method’ folder.

Three different datasets have been used to test this pipeline, one of which has been preserved as an example datasets for the vignettes of this pipeline. More information on the other datasets can be viewed through the publications listed below.

The PBMC dataset is a time series experiment with three time points that explores and compares three AICDA/AID (Activation Induced Cytidine DeAminase) activation cocktails. The experiment seeked to identify which activation cocktail properly activated AID through both the expression of the AID gene and the activation of class switch recombination.

The MURINE dataset is another AID related time series experiment, where mice were generated and one was given an AID activation cocktail. The experiment contains one replicate per condition and 10 timepoints. Due to only having one replicate per condition per timepoint, subsequent timepoints were combined to create an intermediary timepoint with two replicates in each condition.

The Celegans dataset is an ageing related experiment. Two celegans models were used: BY372 and N2. The first being a parkinsons model and the second being a wild type. Each model was split into two groups, one that would be fed krill oil and the other not (control). Each experiment has three timepoints: Days 1, 3, and 6.

A tutorial using the PBMC data can be found within the pipelines website. Additionally the documentation can be found in the references tab.

In order for the pipeline to run properly, a sample file must be provided. Certain requirements ust be met. Column names should be as indicated below, all spaces should be avoided, timepoints should be numerical. It is also important that the sample column contains the same names as the count files provided. Lastly, replicates should always be followed by a ’_#’ as this numbering will be utilized for the organization of the replicates within certain visualizations. Note that in the below example the three files have the same name (ex: C_1). That is to represent the ‘C_1’ sample at its three different time points.

TiSA utilizes DESeq2 to process RNAseq data. It therefore requires that the data be in a raw count format. The pipeline can take in multiple count files within a folder provided that the file names match the sample names given in the sample file. These individual files should be tab delineated, contain only two columns (gene ID and associated count), there should be no column header. Additionally each file should end with a ‘.counts’.

As an alternative, all counts may be within the same text file. In this case the first column should be labeled as ‘gene_id’, and each subsequent column should contain the same sample name as the names found in the sample file. This file should end with ‘.txt’.

In either case, the given information should be counts (round numbers).

For microarray data, a streamlined method is in the works. Currently microarray data needs to be inputed as a Elist, specifically an E list saved as a rds object.

my_path_data<-'data/micro_arr/my_limma_dta.rds'
my_path_sample_dta<-'data/micro_arr/sample_file.csv'

#Set-up time series object parameters
diff_exp_type<-'limma' 

The rds file given to ‘my_path_data’ contains the Elist produced by limma processing of microarray data. It is also important to set the differential gene expression type (diff_exp_type) to ‘limma’.

NOTE: You must still install the TimeSeriesAnalysis package as defined above

Since the pipeline’s source code is available, individuals are free to take the code and adapt it as needed. However for users who would like the easiest approach to use the pipeline as is, it is recommended to clone or download this github repository to your local computer and work from the Rmarkdown method folder within the repository. Cloning or downloading can be done from the TimeSeriesAnalysis github page by using the green button called ‘code’. From there, the Rmarkdown file can be edited.

To edit the file, it is recommended to use rstudio, which can be downloaded here. Note that it is recommended to download the free version of RStudio Desktop, not RStudio Server.

The only code block of code is the parameter set-up block (2nd) as well as the title at the very top of the Rmarkdown document.

The pipeline can be launched from rstudio itself by using the ‘knit’ button, or it can be launched from a command line using the following command from the TimeSeriesAnalysis repository:

Rscript -e "rmarkdown::render('rmarkdown_method/TS_analysis.Rmd',output_file='TS_analysis.html')"

A manuscript detailing the TimeSeriesAnalysis pipeline published with NAR genomics and bioinformatics. TiSA: TimeSeriesAnalysis—a pipeline for the analysis of longitudinal transcriptomics data The version/DOI of this pipeline used within the manuscript is 10.5281/zenodo.7616032.

A ageing related study was performed in part through the use of TiSA. Krill oil protects dopaminergic neurons from age-related degeneration through temporal transcriptome rewiring and suppression of several hallmarks of aging

A longitudinal COVID-19 study was performed using TiSA. Longitudinal whole blood transcriptomic analysis characterizes neutrophil activation and interferon signaling in moderate and severe COVID-19

The ‘clusterGenomics’ package was no longer maintained on CRAN and therefore the necessary scripts were brought over to this package for it’s implementation without the need to download the clusterGenomics package from source. Within this package, in the ‘R/clusterGenomics.R’ code was written by the authors of clusterGenomics. It was imported to TimeSeriesAnalysis by the authors of TimeSeriesAnalysis.

If you have encountered an error or require assistance running the pipeline, feel free to open a github issue, or email me at [email protected]