--- title: "Performing gene set enrichment analyses with sparrow" author: "Steve Lianoglou" date: "`r BiocStyle::doc_date()`" package: "`r BiocStyle::pkg_ver('sparrow')`" abstract: > The sparrow package facilitates the use of gene sets in the analysis of high throughput genomics data. It provides simple execution and comparison of several GSEA approaches through a unified interface within the user's workspace or interactively via a shiny application provided by the sparrow.shiny package. This package also provides an easy wrapper to single sample gene set scoring and geneset-centric heatmaps for visualization. sparrow package version: `r packageVersion("sparrow")` output: BiocStyle::html_document: highlight: pygments toc: true fig_width: 5 css: mystyles.css vignette: > %\VignetteIndexEntry{Performing gene set enrichment analyses with sparrow} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding[utf8]{inputenc} --- ```{r init, include=FALSE, echo=FALSE, message=FALSE, warning=FALSE} knitr::opts_chunk$set( echo=TRUE, warning=FALSE, message=FALSE, error=FALSE) #, dpi=150) # TODO: # 1. update text to favor overrepresentation analysis method `ora` # 2. describe how its data.frame input works (ora) # 3. provide a brief comparison to its performance vs goseq, with and without # accounting for bias). You can take the code for that from the # enrichtest/goseq PAC unit test. ``` # Overview The `{sparrow}` package was built to facilitate the use of gene sets in the analysis of high throughput genomics data (primarily RNA-seq). It does so by providing these top-line functionalities: * The `seas` function is a wrapper that orchestrates the execution of any number of user-specified gene set enrichment analyses (GSEA) over a particular experimental contrast of interest. This will create a `SparrowResult` object which stores the results of each GSEA method internally, allowing for easy query and retrieval. * A sister [`{sparrow.shiny}`][sparrowshiny] package provides an `explore` function, which is invoked on `SparrowResult` objects returned from a call to `seas`. The shiny application facilitates interactive exploration of these GSEA results. This application can also be deployed to a shiny server and can be initialized by uploading a serialized `SparrowResult` `*.rds` file. * An "over representation analysis" method `ora()` which wraps the biased enrichment functionality found within `limma::kegga` and generalizes it to work against data.frame inputs with arbitrary genesets. * The `scoreSingleSamples` function is a wrapper that enables the user to generate single sample gene set scores using a variety of different single sample gene set scoring methods. * Convenience gene set collection retrieval functions that return `BiocSet`s from widely used databases, like `getMSigCollection()` ([MSigDB][msigdb]), `getKeggCollection()` ([KEGG][KEGG]), `getPantherCollection()` ([PANTHER database](pantherdb)), and `getReactomeCollection()` ([Reactome][reactome]) with support for different organisms and identifier types (partially). The initial GSEA methods that sparrow wrapped were the ones provided by limma and edgeR. As such, many analyses using sparrow expect you to re-use the same data objects used for differential expression analysis, namely: * Expression data (an `EList`, `DGEList`, or expression matrix) * A design matrix * A contrast vector/matrix (if your design and comparison require it) Other methods only require the user to provide a ranked vector of statistics that represent some differential expression statistic per gene, and the GSEA is performed by analyzing the ranks of genes within this vector. The user can invoke one `seas()` call that can orchestrate multiple analyses of any type. Currently supported gene set enrichment methods include: ```{r sparrow-methods, echo=FALSE, message=FALSE, warning=FALSE} dplyr::select(sparrow::sparrow_methods(), method, test_type, package) ``` When using these methods in analyses that lead to publication, please cite the original papers that developed these methods and cite sparrow when its functionality assisted in your interpretation and analysis. The sparrow package provides a small example expression dataset extracted from the TCGA BRCA dataset, which is available via the `exampleExpressionSet` function. In this vignette we will explore differential expression and gene set enrichment analysis by examining differences between basal and her2 PAM50 subtypes. # Standard Workflow Let's begin by setting up our work environment for exploratory analysis using the sparrow package. ```{r init-env, warning=FALSE, message=FALSE} library(sparrow) library(magrittr) library(dplyr) library(ggplot2) library(ComplexHeatmap) library(circlize) library(edgeR) library(data.table) theme_set(theme_bw()) ```
Internally, sparrow leverages the [data.table](https://CRAN.R-project.org/package=data.table) package for fast indexing and manipulation over data.frames. All functions that return data.frame looking objects back have converted it from an data.table prior to return. All such functions take an `as.dt` argument, which is set to `FALSE` by default that controls this behavior. If you want `{sparrow}` to return a data.table back to you from some function, try adding an `as.dt = TRUE` argument to the end of the function call.
## Data Setup sparrow is most straightforward to use when our data objects and analysis are performed with either the edgeR or voom/limma pipelines and when we use standard gene identifiers (like esnemble) as `rownames()` to these objects. The `exampleExpressionSet` function gives us just such an object. We call it below in a manner that gives us an object that allows us to explore expression differences between different subtypes of breast cancer. ```{r data-setup, eval=!exists('y.all'), results='hide'} vm <- exampleExpressionSet(dataset = "tumor-subtype", do.voom = TRUE) ``` Below you'll find the `$targets` data.frame of the voomed `EList` ```{r} vm$targets %>% select(Patient_ID, Cancer_Status, PAM50subtype) ```
Note that there are many tutorials online that outline how to generate expression matrices for use with differential expression and analysis, such as the one that is returned from the `exampleExpressionSet` function. Summarizing assay data into such a format is out of scope for this vignette, but you can reference the [airway vignette](http://bioconductor.org/packages/release/data/experiment/vignettes/airway/inst/doc/airway.html) for full details (among others).
## Data Analysis We will identify the genes and genesets that are differentially expressed between the basal and her2 subtypes. The `vm` object has already been `voom`d using this design: ```{r} vm$design ``` We can test for differences between basla and her2 subtypes using the following contrast: ```{r contrast-setup} (cm <- makeContrasts(BvH=Basal - Her2, levels=vm$design)) ``` ### Differential Gene Expression In this section, we first show you the straightforward analysis you would do if you were only testing for differential gene expression. With the data we have at hand, you would simply do the following: ```{r dge-analysis} fit <- lmFit(vm, vm$design) %>% contrasts.fit(cm) %>% eBayes tt <- topTable(fit, 'BvH', n=Inf, sort.by='none') ``` ### Gene Set Enrichment Analysis Given that we now have all of the pieces of data required for a differential expression analysis, performing GSEA is trivial using the `seas` wrapper function. We simply need to now define (1) the battery of gene sets we want to test against, and (2) the GSEA methods we want to explore. ### Gene Sets to Test The *sparrow* package provides a `GeneSetDb` class to store collections of gene sets. The `GeneSetDb` object is used heavily for the internal functionality of `{sparrow}`, however you can provide sparrow with collections of gene sets using other containers from the bioconductor universe, like a `BiocSet::BiocSet` or a `GSEABase::GeneSetCollection`. This package provides convenience methods to convert between these different types of gene set containers. Please refer to *[The GeneSetDb Class](#the-genesetdb-class)* section for more details. The {sparrow} package also provides convenience methods to retrieve gene set collections from different sourckes, like [MSigDB][msigdb], [PANTHER][pantherdb], [KEGG][], etc. These methods are named using the following pattern: `getCollection()` to return a `BiocSet` with the gene sets from the collection, or `getGeneSetDb()` to get a `GeneSetDb` of the same. We'll use the `getMSigGeneSetDb` convenience function provided by the *sparrow* package to load the [hallmark][hallmark] (`"h"`) and [c2 (curated)][c2] (`"c2"`) gene set collections from [MSigDB][msigdb]. ```{r build-gdb} gdb <- getMSigGeneSetDb(c("h", "c2"), "human", id.type = "entrez") ``` :::note To retrieve a `BiocSet` of these same collections, you could do: ```r bsc <- getMSigCollection(c("h", "c2"), "human", id.type = "entrez") ``` ::: You can view a table of the gene sets defined inside a `GeneSetDb` (`gdb`)object via its `geneSets(gdb)` accessor: ```{r geneSets-accessor} geneSets(gdb) %>% head %>% select(1:4) ``` ## Running sparrow Performing multiple gene set enrichment analyses over your contrast of interest simply requires you to provide a `GeneSetDb` (or `BiocSet`) object along with your data and an enumeration of the methods you want to use in your analysis. The call to `seas()` will perform these analyses and return a `SparrowResult` object which you can then use for downstream analysis. ```{r run-multi-GSEA, results='hide', warning=FALSE} mg <- seas( vm, gdb, c('camera', 'fry', 'ora'), design = vm$design, contrast = cm[, 'BvH'], # these parameters define which genes are differentially expressed feature.max.padj = 0.05, feature.min.logFC = 1, # for camera: inter.gene.cor = 0.01, # specifies the numeric covariate to bias-correct for # "size" is found in the vm$genes data.frame, which makes its way to the # internal DGE statistics table ... more on that later feature.bias = "size") ``` We will unpack the details of the `seas()` call shortly ... ## Implicit Differential Expression First, let's note that in addition to running a plethora of GSEA's over our data we've also run a standard differential expression analysis. If you've passed a `matrix`, `ExpressionSet` or `EList` into `seas()`, a *limma*-based `lmFit %>% (eBayes|treat) %>% (topTable|topTreat)` pipeline was run. If a `DGEList` was passed, then `seas` utilizes the *edgeR*-based `glmQLFit %>% (glmQLFTest | glmTreat) %>% topTags` pipeline. The result of the internally run differential expression analysis is accessible via a call to `logFC` function on the `SparrowResult` object: ```{r logFC-results} lfc <- logFC(mg) lfc %>% select(symbol, entrez_id, logFC, t, pval, padj) %>% head ``` We can confirm that the statistics generated internally in `seas()` mimic our explicit analysis above by verifying that the t-statistics generated by both approaches are identical. ```{r compare-dge-t-stats} comp <- tt %>% select(entrez_id, logFC, t, pval=P.Value, padj=adj.P.Val) %>% inner_join(lfc, by='entrez_id', suffix=c('.tt', '.mg')) all.equal(comp$t.tt, comp$t.mg) ``` The internally performed differential expression analysis within the `seas()` call can be customized almost as extensively as an explicitly performed analysis that you would run using limma or edgeR by sending more parameters through `seas()`'s `...` argument. See the *[Custom Differential Expression](#custom-differential-expression)* section further in the vignette as well as the help available in `?calculateIndividualLogFC` (which is called inside the `seas()` function) for more information. ## Explicit GSEA We also have the results of all the GSEA analyses that we specified to our `seas` call via the `methods` parameter. ```{r mg-res} mg ``` The table above enumerates the different GSEA methods run over each geneset collection in the rows. The columns enumerate the number of genesets that the collection has in total (`geneset_count`), and how many were found significant at a given FDR, which is set to 20% by default. The `show` command for the `SparrowResult` object simply calls the `tabulateResults()` function, which you can call directly with the value of `max.p` that you might find more appropriate. ## Exploring Results GSEA results can be examined interactively via the command line, or via a shiny application. You can use the `resultNames` function to find out what GSEA methods were run, and therefore available to you, within the the `SparrowResult` object: ```{r resultnames} resultNames(mg) ``` Note that when running an "over representation analysis" `"ora"` (or `"goseq"`), it will be run three different ways. The tests will be run first by testing all differentially expressed genes that meet a given set of min logFC and max FDR thresholds, then separately for only genes that go up in your contrast, and a third time for only the genes that go down. The individual gene set statistics generated by each method are available via the `result` function (or several can be returned with `results`): ```{r gsea-res} cam.res <- result(mg, 'camera') cam.go.res <- results(mg, c('camera', 'ora.up')) ``` You can identify genesets with the strongest enrichment by filtering and sorting against the appropriate columns. We can, for instance, identify which hallmark gene sets show the strongest enrichment as follows: ```{r camera-summary} cam.res %>% filter(padj < 0.1, collection == 'H') %>% arrange(desc(mean.logFC)) %>% select(name, n, mean.logFC, padj) %>% head ``` You can also list the members of a geneset and their individual differential expression statistics for the contrast under test using the `geneSet` function. ```{r geneset-result} geneSet(mg, name = 'HALLMARK_WNT_BETA_CATENIN_SIGNALING') %>% select(symbol, entrez_id, logFC, pval, padj) %>% head() ```
The results provided in the table generated from a call to `geneSet` are independant of GSEA method. The statistics appended to the gene set members are simply the ones generated from a differential expression analysis.
## Plotting `{sparrow}` provides a number of interactive plotting facilities to explore the enrichment of a single geneset under the given contrast. In the boxplots and density plots shown below, the log fold changes (logFCs) (or t-statistics) for all genes under the contrast are visualized in the "background" set, and these same values are shown for the desired geneset under the "geneset" group. The logFC (or t-statistics) of the genes in the gene set are plotted as points, which allow you to hover to identify the identity of the genes that land in the regions of the distributions you care about.
Including interactive plots increases the size of the vignette's by *a lot* and will be rejected by the bioconductor build servers, so all plots included in this vignette are static snapshots of the javascript enabled plots you would normally get from `iplot()`.
**Boxplot** ```{r iplot-wnt-beta, fig.asp=1, eval=FALSE} iplot(mg, 'HALLMARK_WNT_BETA_CATENIN_SIGNALING', type = 'boxplot', value = 'logFC') ``` ![boxplot of geneset log2FC's](images/static-iplot-wnt-beta-box.png) **Density** ```{r iplot-wnt-beta-density, fig.asp=1, eval=FALSE} iplot(mg, 'HALLMARK_WNT_BETA_CATENIN_SIGNALING', type = 'density', value = 'logFC') ``` ![density plot of geneset log2FC's](images/static-iplot-wnt-beta-density.png) **GSEA plot** ```{r iplot-wnt-beta-gsea, fig.asp=1, eval=FALSE, message=FALSE, warning=FALSE} iplot(mg, 'HALLMARK_WNT_BETA_CATENIN_SIGNALING', type = 'gsea', value = 'logFC') ``` ![gsea plot of geneset log2FC's](images/static-iplot-wnt-beta-gsea.png) ### Interactive Exploration A sister [`{sparrow.shiny}`][sparrowshiny] package is available that can be used to interactively explore `SparrowResult` objects to help you try to make sense of the enrichment hits you get (or not!). The application can be invoked as follows: ```r sparrow.shiny::explore(mg) ``` ![Screen shot of interactive sparrow exploration](images/sparrow-shiny.png) Please refer to the `"sparrow-shiny"` vignette in the [`{sparrow.shiny}`][sparrowshiny] package for documentation on the application's use. :::note The `{sparrow.shiny}` package is currently only available to install from GitHub, but will be available through Bioconductor soon. ::: # Singe Sample Gene Set Scoring It can be both convenient and effective to transform a gene-by-sample expression matrix to a geneset-by-sample expression matrix. By doing so, so we can quickly identify biological processes that are up/down regulated (loosely speaking) in each sample. We can generate single sample gene set scores using the gene sets defined in a `GeneSetDb` using the `scoreSingleSamples` function. This function takes a `GeneSetDb`, an expression container, and a `methods` argument, which is analagous to the `methods` argument in the `seas()` call: it defines all of the scoring methos the user wants to apply to each sample. Let's pick a few gene sets to score our samples with for this exercise. We'll take the significant hallmark gene sets, or any other significant gene set that has a large (on average) log fold change between conditions. ```{r ssgenesets} sig.res <- cam.res %>% filter(padj < 0.05 & (grepl("HALLMARK", name) | abs(mean.logFC) >= 2)) gdb.sub <- gdb[geneSets(gdb)$name %in% sig.res$name] ```
Refer to the [Subsetting a GeneSetDb](#subsetting-a-genesetdb) section to learn how to subset a `GeneSetDb` object to create a derivative object with fewer gene sets.
Recall that the GSEA analysis we performed was perfomed between the Basal and Her2 subtypes, so we will use an expression matrix that only has the samples from those two groups. ```{r subset-exprs} vm.bh <- vm[, vm$targets$PAM50subtype %in% c("Basal", "Her2")] ``` ## Generating Single Sample Gene Set Scores Once we have a `GeneSetDb` object that contains all of the gene sets we wish to use to create single sample gene set scores, we can use the `scoreSingleSamples` function to produce these scores using a variety of algorithmes, which the user species using the `methods` parameter. The `scoreSingleSamples` function will return a long `data.frame` with `length(methods) * ncol(exprs)` rows. Each row represents the score for the given `sample` using the specified `method`. You can subset against the `method` column to extract all of the single sample scores for a given method. ```{r ssscore, warning=FALSE} scores <- scoreSingleSamples(gdb.sub, vm.bh, methods = c('ewm', 'ssgsea', 'zscore'), ssgsea.norm = TRUE, unscale=FALSE, uncenter=FALSE, as.dt = TRUE) ``` We can see how the scores from different methods compare to each other: ```{r sss-pairs, warning=FALSE} # We miss you, reshape2::acast sw <- dcast(scores, name + sample_id ~ method, value.var="score") corplot(sw[, -(1:2), with = FALSE], cluster=TRUE) ``` It is, perhaps, interesting to compare how the `ewm` method scores change when we choose not to "uncenter" and "unscale" them: ```{r, warning=FALSE} ewmu <- scoreSingleSamples(gdb.sub, vm.bh,methods = "ewm", unscale = TRUE, uncenter = TRUE, as.dt = TRUE) ewmu[, method := "ewm_unscale"] scores.all <- rbind(scores, ewmu) swa <- dcast(scores.all, name + sample_id ~ method, value.var="score") corplot(swa[, -(1:2), with = FALSE], cluster=TRUE) ``` Further exposition on the "ewm" (eigenWeightedMean) scoring method can be found in the `?eigenWeightedMean` function. ## Visualizing Single Sample Gene Set Scores The "long" data.frame nature of the results produced by `scoreSingleSamples` makes it convenient to use with graphing libraries like ggplot2 so that we can create arbitrary visualizations. Creating boxplots for gene sets per subtype is an easy way to explore these results. Let's annotate each row in `scores.all` with the subtype annotation and observe how these methods score each sample for a few gene sets. ```{r anno-scores} all.scores <- scores.all %>% inner_join(select(vm.bh$targets, sample_id=Sample_ID, subtype=PAM50subtype), by = "sample_id") some.scores <- all.scores %>% filter(name %in% head(unique(all.scores$name), 5)) ggplot(some.scores, aes(subtype, score)) + geom_boxplot(outlier.shape=NA) + geom_jitter(width=0.25) + facet_grid(name ~ method) ``` # Gene Set Based Heatmap with mgheatmap We often want to create expression based heatmaps that highlight the behavior of gene sets across our samples. The `mgheatmap` function uses the [ComplexHeatmap][cplxhmap] package to create two different types of heatmaps: 1. Gene based heatmaps, that split the genes (rows) based on their genesets 2. Single sample gene set based heatmaps, optionally split by gene set collection. The `mgheatmap` function has a set of arguments that customize how the heatmap is to be created (gene level vs. gene set level, whether to split it, etcv) and will also use the `...` argument to pass any parameters down to the inner `ComplexHeatmap::Heatmap` function call and customize its behavior. The `mgheatmap` function returns a `ComplexHeatmap,Heatmap` object for plotting or combining with other ComplexHeatmap heatmaps or annotations in order to create arbitrarily complex/informative heatmap figures. [cplxhmap]: http://bioconductor.org/packages/ComplexHeatmap ## Gene level based heatmap (from genesets) You can plot a heatmap of the genes from a predefined set of gene sets by providing the gene sets you want to visualize in a `GeneSetDb` object. We'll create a new `GeneSetDb` object using the first two gene sets in `gdb.sub` and draw a heatmap of their expression. ```{r gheatmap, fig.height=8, fig.width=4} gs.sub <- geneSets(gdb.sub) gdb.2 <- gdb.sub[geneSets(gdb.sub)$name %in% head(gs.sub$name, 2)] col.anno <- HeatmapAnnotation( df = vm.bh$targets[, 'PAM50subtype', drop = FALSE], col = list(PAM50subtype = c(Basal = "gray", Her2 = "black"))) mgheatmap(vm.bh, gdb.2, aggregate.by = "none", split = TRUE, show_row_names = FALSE, show_column_names = FALSE, recenter = TRUE, top_annotation = col.anno, zlim = c(-3, 3)) ``` ## Gene set-based heatmap You can often get a higher information:ink ratio by plotting heatmaps based on single sample gene set scores as opposed to the genes that make up a geneset. Let's see what the simple 2-geneset version of the heatmap above looks like: ```{r gshm2, fig.height = 2.5, fig.width = 8} mgheatmap(vm.bh, gdb.2, aggregate.by = "ewm", split = FALSE, show_row_names = TRUE, show_column_names = FALSE, top_annotation = col.anno) ``` Plotted in this way, we can now show the activity of a greater number of genesets ```{r gshm-all, fig.height = 6, fig.width = 8} mgheatmap(vm.bh, gdb.sub, aggregate.by = 'ewm', split=TRUE, recenter = TRUE, show_row_names=TRUE, show_column_names=FALSE, top_annotation=col.anno, zlim = c(-2.5, 2.5)) ``` # The GeneSetDb Class The GeneSetDb class was developed to address the internal needs of the sparrow package for fast look up, subsetting, cross reference, etc. of a collection of gene sets. At the time (~2015), it was developed because the classes used for this purpose in the bioconductor ecosystem (a `GSEABase::GeneSetCollection`, or a simple list of gene vectors) didn't cut the mustard. More recently, bioc-core has developed a new class called a `BiocSet` that is feature-rich and shares significant overlap with the features in the `sparrow::GeneSetDb` class. Although we can't quite replace the internals of {sparrow} to use the `BiocSet` just yet, users are encouraged to provide collections of gene sets in the form of a `BiocSet` everywhere {sparrow} functions require gene set collections, like `seas()` and `scoreSingleSamples()`. You can also convert a `sparrow::GeneSetDb()` to a `BiocSet` via a simple call: `as(gdb, "BiocSet")`. The remainder of this section provides a quick overview of the `GeneSetDb` class. The GeneSetDb object uses the `data.table` package internally for fast lookup. Internally the collection of gene set information is minimally stored as a three-column `data.table` in "long form", which has the following columns: * collection * name * feature_id More columns can be added to the internal `data.table` (a "symbol" column, for instance), but those are the only three you need. To see what we are talking about, exactly, you can call the `as.data.frame` function on a `GeneSetDb` object: ```{r} as.data.frame(gdb)[c(1:5, 201:205),] ``` The `(collection,name)` tuple is the primary key of a gene set. The `feature_id` column stores gene identifiers. For the time being, it will be most natural for these IDs to simply be ensembl gene identifiers (or entrez ids) as many of the annotation databases use these identifiers, as well. In reality, you will want the values in the `feature_id` columns to match with the feature id's you have in your data container (ie. the `rownames()` of a `SummarizedExperiment`, for instance). ## Building a GeneSetDb The sparrow package provides convenience functions to fetch genesets from many sources and convert them into a GeneSetDb object. The two most useful sources may be: * [MSigDB][msigdb] via `getMSigGeneSetDb(...)`. Although the core `sparrow` package provides the getter function for these genesets, the main data retrieval functionality is provided through the [msigdbr][] package. * [PANTHER][pantherdb] (pathways and GOSLIM) via `getPantherGeneSetDb()` * [KEGG][] via `getKeggGeneSetDb(...)` We also provide similarly named methos to retrieve these gene set collections as a `BiocSet`, just substitute `"Collection"` for `"GeneSetDb"`, ie. `getMsigCollection(...)`, `getPantherCollection(...)`, and `getKeggCollection(...)`. You can create a custom `GeneSetDb` via the `GeneSetDb()` constructor, which accepts the following types of inputs. 1. A `BiocSet` 2. A `GeneSetCollection` 3. A data.frame of geneset membership. This requires `collection`, `name`, and `feature_id` columns. Reference the output of `as.data.frame(gdb)` shown above. 4. A named list of gene identifier vectors that represent genesets for a single collection 5. A named list of (2)-like lists. The top level names are the names of the different collections, and each sublist represents the genesets in that collection. Two `GeneSetDb` objects can be combined using the `cobine()` function. For now it is your responsibility to ensure that the two `GeneSetDb` objects are "reasonably conformable", ie. they use the same types of gene identifiers, and are referencing the same species, etc. ```{r, eval=FALSE} msigdb <- getMSigGeneSetDb('H', 'human') goslimdb <- getPantherGeneSetDb('goslim', 'human') gdb.uber <- combine(msigdb, goslimdb) ``` See the help and examples in `?GeneSetDb` for more information.
For some reason the `PANTHER.db` package needs to be installed in a user-writable package location for this to work properly. If you see an error that speaks to using "rsqlite to write to a readonly database", you will have to re-install `PANTHER.db` in a user-writable directory using `BiocManager::install("PANTHER.db")`
## Subsetting a GeneSetDb The subsetting functionality for a `GeneSetDb` is a bit clunky. We assume you want to subset a GeneSetDb to include a subset of, well, gene sets. One way you can do that is to provide a logical vector that is as long as there are gene sets in the GeneSetDb as an index. For instance, if we want to include only the genesets in [CP:PID][msigdbpid], you can do that. This subcatory information is stored in the `"subcategory"` column from `geneSets(gdb)` ```{r subset-gdb-by-metadata} keep <- geneSets(gdb)$subcategory == "CP:PID" gdb.sub <- gdb[keep] geneSets(gdb.sub) %>% head() ``` You can also subset a `GeneSetDb` to only include gene sets that contain certain features: ```{r subset-gdb} gdb.sub2 <- subsetByFeatures(gdb, c('10014', '1454')) nrow(gdb); nrow(gdb.sub2) ``` ## Active vs Inactive Gene Sets A `GeneSetDb` is used to hold "the universe" of genes that belong to different gene sets across different collections. Depending on the assay performed to measure these genes, the set of genes you observe in your study will likely be a subset of the genes in the `GeneSetDb`. As such, prior to using a `GeneSetDb` for GSEA, it must be "conformed" to a target object that will be used for the input to the GESA (either a matrix of expression, or a pre ranked vector of statistics). This step will index into the target expression object and identify which rows of the object correspond to which genes in the `GeneSetDb`. "Conformation" happens automatically within the `seas()` call, but we call it explicitly below to outline its functionality. The command below conforms the `GeneSetDb` to our target "voomed" `EList`, and deactivates gene sets (i.e. removes them from downstream GSEA) that have less than 10 or more than 100 genes that were found in `vm`: ```{r} gdbc <- conform(gdb, vm, min.gs.size=10, max.gs.size=100) head(geneSets(gdbc, active.only=FALSE)) ``` We can see that, only 23 of the 26 genes in the `(C2,ABBUD_LIF_SIGNALING_1_DN)` were found in the rows of `vm`, and the `(C2,ABBUD_LIF_SIGNALING_2_DN)` was "deactivated." Deactivated (`active == FALSE`) gene sets will be ignored during downstream analyses. This gene set was deactivated because it only has five "conformed" genes, but the minimum geneset size we wanted to consider (`min.gs.size`) was set to ten in our call to `conform`. ## Accessing members of a gene set The `geneSet` and `featureIds` functions allow the user to identify the genes found in a gene set. Both of these functions take an `active.only` argument, which is `TRUE` by default. This specifies that only the genes that have been successfully conformed to a gene set should be the ones that are returned. For instance, we can identify which genes belong to the `(C2,ABBUD_LIF_SIGNALING_1_DN)`, and which three were not found in `vm` like so: ```{r} missed <- setdiff( featureIds(gdbc, 'C2', 'ABBUD_LIF_SIGNALING_1_DN', active.only=FALSE), featureIds(gdbc, 'C2', 'ABBUD_LIF_SIGNALING_1_DN', active.only=TRUE)) missed ``` or we can use the `geneSet` function to return a `data.frame` of these results: ```{r} gdbc %>% geneSet('C2', 'ABBUD_LIF_SIGNALING_1_DN', active.only = FALSE) %>% subset(feature_id %in% missed) ``` ## Mapping of gene set featureIds to target expression containers It may be that the IDs used in a gene set collection are different from the ones used as the rownames of your expression container. For instance, the IDs used for a given gene set collection in the `GeneSetDb` might be Ensembl gene identifiers, but the rownames of the expression object might be Entrez ID. This is where the `mapping` parameter becomes useful. The `GeneSetDb` class has a concept of an internal `featureIdMap` to accommodate these scenarios, which would allow for a non-destructive mapping of the original IDs to a new "ID space" (entrez to ensembl, for instance). This functionality is not ready for this release, but it's just a note to keep the user aware of some future development of the package. For the time being, the user is required to manually map the feautreIds in their expression matrix to be concordant with the ones found in the `GeneSetDb`. **In the meantime**, a `renameRows` convenience function is provided here to easily rename the rows of our expression container to different values. For instance, to rename this is how you might rename the rows of your assay container to use symbols: ```{r} vm <- exampleExpressionSet() vms <- renameRows(vm, "symbol") head(cbind(rownames(vm), rownames(vms))) ``` We grabbed the `symbol` column from `vm$genes` and "smartly" renamed the rows of `vm` with the values there. Refer to the `?renameRows` man page for more details. This, of course, still requires you to manually fetch and map identifiers, but still ... # Customizing Analyses The internal differential expression analysis as well the gene set enrichment analyses can be customized by passing parameters through the `...` in the `seas()` function. ## Custom Differential Expression The internal differential expression pipeline, exported via the `calculateIndividualLogFC` function allows the end user to configure an "arbitrarily complex" differential expression analysis using either edgeR's quasilikelihood framework (if the input is a DGEList) or a direct limma analysis (with a pre-voomed EList, expression matrix, or whatever). User's should refer to the `?calculateIndividualLogFC` help page to see which parameters are exposed for a differential expression analysis and configure them accordingly. When calling `seas()` use these same parameters in the call and they will be provided to `calculateIndividualLogFC`. For instance, if you wanted to use limma's "treat" functionality to specify a minimal log fold change threshold for statistical significance, you would do so as follows: ```{r, eval=FALSE} mg <- seas(vm, gdb, "goseq", design = vm$design, cm[, 'BvH'], treat.lfc=log2(1.5), ## feature length vector required for goseq feature.bias=setNames(vm$genes$size, rownames(vm))) ``` Using the internal `treat` functionality would really only affect enrichment tests that first threshold the genes in your experiment as "significant" or not, like `goseq` and not tests like `camera`. ## Custom GSEA The GSEA methods that are wrapped by `seas()` all take the same parameters that are defined by their implementation. Simply pass these parameters down via the `...` in the `seas()` call. For instance, you can read `?camera` to find that the `camera` method accepts an `inter.gene.cor` parameter, and `?roast` will tell you that you can specify the number of rotations used via the `nrot` parameter. ```{r, eval=FALSE} mgx <- seas(vm, gdb, c('camera', 'roast'), design = vm$design, contrast = cm[, 'BvH'], inter.gene.cor = 0.04, nrot = 500) ``` # Reproducibility
Session Information ```{r session-info} sessionInfo() ```
[BiocSet]: http://bioconductor.org/packages/release/bioc/html/BiocSet.html [msigdb]: http://software.broadinstitute.org/gsea/msigdb/ [msigdbr]: https://cran.r-project.org/package=msigdbr [msigdbpid]: https://www.gsea-msigdb.org/gsea/msigdb/genesets.jsp?collection=CP:PID [hallmark]: http://www.cell.com/cell-systems/abstract/S2405-4712(15)00218-5 [isigdb]: http://www.cell.com/immunity/abstract/S1074-7613(15)00532-4 [c2]: http://software.broadinstitute.org/gsea/msigdb/collections.jsp#C2 [c5]: http://software.broadinstitute.org/gsea/msigdb/collections.jsp#C5 [c7]: http://software.broadinstitute.org/gsea/msigdb/collections.jsp#C7 [KEGG]: https://www.genome.jp/kegg/ [pantherdb]: http://pantherdb.org [reactome]: https://reactome.org/ [pgoslim]: http://www.pantherdb.org/panther/ontologies.jsp [sparrowshiny]: https://github.com/lianos/sparrow.shiny