The FacileAnalysis package defines a set of analysis tasks over genomic data in a modular fashion, which requires these data to be stored in a container that implements the FacileData API (aka a
The over-arching goal of developing analyses within this framework is to enable quick and effortless interrogation of genomic data by enabling a hybrid interactive and code-driven approach to data analysis. Analyses can be either completely code-driven, GUI driven, or some mix of the two.
To achieve this goal, analysis modules break down a general analysis task into smaller constituent steps, the results of which can:
%>%) to perform a complete analysis;
Please refer to the RNA-seq Analysis: The Facile Way to get an idea of how to use these tools when analyzing RNA-seq data.
The analyses implemented within this package are listed below, with links to vignettes that describe their functionality in more detail:
fpca: Interactive Principal Components Analyses
fdge: Interactive Differential Gene Expression Analysis
fsea: Interactive (Gene) Set Enrichment Analysis
A note on the experimental lifecycle: This package is tagged as “experimental” due to its limited use by a broader audience, and not as a sign of the commitment to its development or how long it has been in (internal) use. As we find the edge cases and pain points in the APIs through broader adoption, we expect to soon move to “maturing” (and eventually “stable”) lifecycle.
We’ll include an exemplar differential expression analysis here in order to touch on some of the guiding principles of an analysis modules mentioned above.
We have first defined a complete differential expression analysis (
fdge), by breaking it down into the following steps:
FacileDataStoreto perform the analysis over;
The example below identifies genes differentially expression between tumor and normal samples (and controlling for sex) in the “BLCA” indication of the example TCGA dataset included in the FacileData package.
library(FacileData) library(FacileAnalysis) efds <- exampleFacileDataSet() # Step 1: define the samples implicated in our test samples <- filter_samples(efds, indication == "BLCA") # Step 2: define the model (design) for the test model <- flm_def(samples, covariate = "sample_type", numer = "tumor", denom = "normal", batch = "sex") # Step 3: configure the options to run the test, which include the assay that # holds the data used to test and the statistical method/framework we # should use to perform the test vdge <- fdge(model, assay_name = "rnaseq", method = "voom")
Perhaps you prefer the edgeR/QLF analysis framework, instead? No problem, we only need to tweak one of the parameters in the last step of the pipeline:
qdge <- fdge(model, assay_name = "rnaseq", method = "edgeR-qlf")
… or DESeq2, perhaps? No problem, we accept pull requests!
There are a number of S3 methods a
FacileAnalysisResult needs to define in order to be complete. The
report() methods allow the end-user to interrogate (and report) the results of an analysis at different levels of interactivity.
Let’s take a look at how these work over a
shine(aresult, ...) method provides the richest interactive view over a
FacileAnalysisResult by launching a shiny gadget that enables the end-user to fully interrogate the results.
While the output of the
viz() functions can be used directly in Rmarkdown reports,
report(aresult, ...) is meant to create a “more complete” (perhaps multi-htmlwidget) view over the result that can be more suitable for inclusion into an Rmarkdown report.
The same differential expression analysis that created the
vdge object above can be performed entirely interactively, with the same results.
We can either start the analysis from the same predefined set of samples, but define the linear model and testing framework to use interactively, by launching a the facile differential gene expression gadget:
vdge2 <- fdgeGadget(samples)
Or we can perform the whole thing via a GUI which lets us select the subset of samples and run the differential expression analysis without using any code at all:
vdge3 <- fdgeGadget(efds)
Assuming the same filtering and testing strategies were selected in the GUI using the
fdgeGadget calls above, the objects they return will all be equivalent to the
vdge result, which was entirely generated programmatically.
# install.packages("BiocManager") BiocManager::install("facilebio/FacileAnalysis")
*.movfiles to animated gifs.