NEWS.md

In lianos/multiGSEA: A unified interface to a plethora of gene set enrichment analysis methods

multiGSEA 1.0.0 (Bioconductor release)

New Features

• Adds preliminary support for data-frame based inputs to multiGSEA() to more easily support pre-ranked and ORA-like tests. When x is a data.frame, use the rank_by parameter to specify the column name that holds the rank metric, and rank_order to specify how to order the ranks ("descending", "ascending", or "ordered").

The implementation is still tied to the original design of package which only supported a numeric ranking vector, and hijacks the whole xmeta. workaround. Consider using FacileAnalysis::ffsea(data.frame, gdb) if you want a more coherent interface.

Breaking Changes

• featureId used all-over-the-place and sample columns from single sample scoring are renamed to feature_id and sample_id to play better with facile.bio

• enrichtest renamed to ora (over representation analysis) to be more inline with nomenclature in the field

multiGSEA 0.10

User Visible Changes

• MSigGeneSetDb updated to v6.1 and data refactored out into GeneSetDb.* packages
• geneSet,MultiGSEAResult method does not remove duplicated columns present in both geneSet(gdb) and logFC, but rather appends a *.gs suffix to the duplicated columns from geneSet. This happened because I defined a geneset with a logFC column (from the experiment the set was derived) from, but this collided with (and superseded) the logFC value for the gene returned from logFC(mgresult), and messed up interpretation of a geneset's move across a contrast.
• Introduces idea of gene- and geneset-level metadata in a GeneSetDb. The former (gene-level metadata) was always possible by creating a GeneSetDb with a data.frame which has more columns than necessary, ie. instead of just having a collection,name,featureId data.frame, you could add a symbol column. This would simply be part of the GeneSetDb@db data.table. If there are columns in your input data.frame that are all the same value per collection,name subset, then these are marked as "geneset level annotations" and tacked onto the GeneSetDb@table data.table. I put this in to pave the way for simple support of "tags" per geneset, but there are other uses as well.

multiGSEA 0.6.x

User Visible Changes

• MSigDB version updated to 5.2 in 0.6.5
• Interactive plots can be generated via iplot
• single sample scoring methods broken out into more modular functions
• svdScore added. Executes Jason's SVD scoring trick and also calculates "all the usual" prcomp stuff.
• Adds a [.GeneSetDb method so we can subset the geneSets out of a GeneSetDb more easily. Would still like to have a fully operational subset(gdb, ...).

New Features

• Adds svdGeneSetTest method. A gene x sample expression matrix is transformed into a geneset x sampel matrix, and "normal" differential expression expression testing is run over the geneset scores.

multiGESA 0.4.x

User Visible Changes

• scoreSingleSamples now defualts to returning a melted data.frame of results (previously we returned a (list of) matrix of scores). melted=FALSE parameter was changed to as.matrix=FALSE

• Internal MSigDB collections updated to v5.1 (as of v0.4.30). The major difference in v5.1 vs v5.0 seems to be an updated c7 (immunologic) collection, which comes on the heels of this paper: Compendium of Immune Signatures Identifies Conserved and Species-Specific Biology in Response to Inflammation http://www.cell.com/immunity/abstract/S1074-7613(15)00532-4

• All methods now default to returning data.frame(s) instead of data.table(s) data.table is still used internally, however I understand that some people don't want to have them returned into their workspace. A global option (multiGSEA.df.return) is available for you to tweak this behavior, ie. set options(multiGSEA.df.return='data.tabe') to have geneSets(gdb) return a data.table instead of a data.frame.

• Changed geneSetFeatureStatistics to geneSetsStats

New Features

• use.treat parameter has been added to relevant places to run the internal differential testing via the "treat" framework made available in limma and edgeR via the treat and glmTreat "pipelines," respectively. By default, all pipelines do not use the treat framework (but perhaps they should).

• goseq, fry, and romer testing methods have been added.

• Support for DGEList expression input fully baked in. Previously, a DGEList was shot through voom to be used internally. Internal differential gene expression statistics on DGELists use edgeR's quasi-likelihood framework. See ?edgeR::glmQLFit for more information, and particulary reference Aaron Lun's tutorial on using this approach here:

  It's DE-licious: a recipe for differential expression analyses of RNA-seq experiments using quasi-likelihood methods in edgeR http://www.statsci.org/smyth/pubs/QLedgeRPreprint.pdf

• Adds support for PANTHER pathway (and GOSLIM) genesets from PANTHER.db package via the getPantherGeneSetDb function.

• Added geneSetFeatureStats to return the logFC (and membership) information for a geneset after a multiGSEA run.

• Added constructors and as() coercion functions to create GeneSetDb's from GeneSetCollection(s) and data.frames (and vice versa).

• scoreSingleSamples is another wrapper function which calls several single sample geneset scoring algorithms, including the ones provided by the GSVA package (such as plage and ssGSEA)

• Adds expression-utils.R for CPM/RPKM and meltx. Also has methods to facilitate "universal access" to pData, fData, and expression data from the different often-used expression containers (ExpressionSet, EList, DGEList, SummarizedExperiment)

• MSigDB definitions were updated to v5.0, which adds a new "hallmark (h)" gene set. A call to getMSigDBset without a version parameter will return the v5.0 data.

Note that MSigDB provides the entrez IDs for each geneset as human genes. The mouse version of these gene sets was constructing by mapping the human entrez ID's to their mouse orthologs via igis::orthologs. When this mapping returns multiple mouse IDs for a single human ID, all of the mouse IDs are kept in the mouse geneset.

To get the previous v4.0 genesets (provided by WEHI), you would simply specify version='v4.0' in the getMSigDBSet like so:

  ```r
  gdb5 <- getMSigDBsetc(c('c2', 'c7'), species='mouse')
  gdb4 <- getMSigDBsetc(c('c2', 'c7'), species='mouse', version='v4.0')
  ````

Bug Fixes

• do.hyperGeometricTest was failing when the expression object x was a matrix.

multiGSEA 0.1.0

New Features

• multiGSEA returns a MultiGSEAResult object which holds all of the things one would need to analyze/plot/poke at the results of the function call.

• Plotting functions added so user can plot results of arbitrary genesets.

User Visible Changes

• This version is a reboots the internals of this package so that its design is cleaner. All of the reporting bits were removed (rmd.plugins handles that now), as well as significant additions to the API of this packages. The changes are too numerous to report here. The previous version before the refactor have been put here for posterity:

• http://resscm.gene.com/bioinfo/projects/R/tags/multiGSEA_pre-refactor

• http://resscm.gene.com/bioinfo/projects/R/tags/rmd.plugins_pre-refactor

lianos/multiGSEA documentation built on Nov. 17, 2020, 1:26 p.m.