Package 'ReactomePA'

Bayesian term selection for Reactome enrichment results

Description

'bayes_reactome()' adds a model-based selection layer on top of 'enrichPathway()' results. It estimates the posterior probability that each candidate Reactome pathway is an active biological program explaining the observed input genes.

Usage

bayes_reactome(
  x,
  candidate = c("top", "significant", "all"),
  n_terms = 200,
  by = "p.adjust",
  prior = 0.1,
  false_positive = 0.01,
  false_negative = 0.1,
  n_iter = 5000,
  burnin = 1000,
  thin = 1,
  posterior_cutoff = 0.5,
  seed = NULL,
  verbose = FALSE
)

Arguments

x	An 'enrichResult' object from 'enrichPathway()'.
candidate	Candidate terms to include. '"significant"' uses 'as.data.frame(x)', '"all"' uses 'x@result', '"top"' uses the top 'n_terms' rows from 'x@result' ordered by 'by'; or provide a character vector of term IDs.
n_terms	Maximum number of candidate terms when 'candidate = "top"' or when more candidates are supplied than this value. Use 'Inf' to disable.
by	Column used to order candidate terms.
prior	Prior probability that a term is active.
false_positive	Probability of observing a gene not covered by active terms.
false_negative	Probability of missing a gene covered by active terms.
n_iter	Total number of MCMC iterations.
burnin	Number of initial iterations discarded.
thin	Keep one sample every 'thin' iterations after burn-in.
posterior_cutoff	Terms with posterior greater than or equal to this value are marked active.
seed	Optional random seed.
verbose	Print sampler progress.

Details

This is a thin wrapper around 'enrichit::bayes_enrich()', intended as a result-compression and interpretation layer for Reactome pathway enrichment. Reactome's hierarchical pathway structure often produces many redundant significant terms — Bayesian selection helps distinguish truly active programs from statistically significant but dependent ones.

Value

The input 'enrichResult' object with additional columns in '@result': 'posterior', 'posterior_odds', 'bayes_rank', 'bayes_active', 'bayes_covered_gene', and 'bayes_covered_count'.

See Also

[enrichit::bayes_enrich()]

Examples

## Not run: 
  gene <- c("11171", "8243", "112464", "2194",
            "9318", "79026", "1654", "65003",
            "6240", "3476", "6238", "3836",
            "4176", "1017", "249")
  x <- enrichPathway(gene, pvalueCutoff = 0.1)
  bx <- bayes_reactome(x)
  head(as.data.frame(bx))

## End(Not run)

---

Datasets sample contains a sample of gene IDs.

Description

Datasets sample contains a sample of gene IDs.

---

Pathway Enrichment Analysis of a gene set. Given a vector of genes, this function will return the enriched pathways with FDR control.

Description

Pathway Enrichment Analysis of a gene set. Given a vector of genes, this function will return the enriched pathways with FDR control.

Usage

enrichPathway(
  gene,
  pvalueCutoff = 0.05,
  pAdjustMethod = "BH",
  universe = NULL,
  weight = NULL,
  minGSSize = 10,
  maxGSSize = 500,
  qvalueCutoff = 0.2,
  organism = "human",
  readable = FALSE
)

Arguments

gene	a vector of entrez gene id.
pvalueCutoff	Cutoff value of pvalue.
pAdjustMethod	one of "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr", "none"
universe	background genes
weight	A named numeric vector of weights for background genes. If provided, Weighted ORA will be performed using Wallenius' noncentral hypergeometric distribution.
minGSSize	minimal size of genes annotated by Ontology term for testing.
maxGSSize	maximal size of each geneSet for analyzing
qvalueCutoff	Cutoff value of qvalue
organism	one of "human", "rat", "mouse", "celegans", "yeast", "zebrafish", "fly".
readable	whether mapping gene ID to gene Name

Value

A enrichResult instance.

Author(s)

Guangchuang Yu http://ygc.name

See Also

enrichResult-class

Examples

## Not run: 
	gene <- c("11171", "8243", "112464", "2194",
			"9318", "79026", "1654", "65003",
			"6240", "3476", "6238", "3836",
			"4176", "1017", "249")
	yy = enrichPathway(gene, pvalueCutoff=0.05)
	head(summary(yy))

## End(Not run)

---

getDb

Description

mapping organism name to annotationDb package name

Usage

getDb(organism)

Arguments

organism

one of supported organism

Value

annotationDb name

Author(s)

Yu Guangchuang

---

gsePathway

Description

Gene Set Enrichment Analysis of Reactome Pathway

Usage

gsePathway(
  geneList,
  organism = "human",
  exponent = 1,
  weight = NULL,
  minGSSize = 10,
  maxGSSize = 500,
  eps = 1e-10,
  pvalueCutoff = 0.05,
  pAdjustMethod = "BH",
  verbose = TRUE,
  nPerm = 1000,
  method = "multilevel",
  adaptive = FALSE,
  minPerm = 101,
  maxPerm = 1e+05,
  pvalThreshold = 0.1,
  seed = FALSE,
  ...
)

Arguments

geneList	order ranked geneList
organism	organism
exponent	weight of each step
weight	A named numeric vector of weights for genes. The names should match the names of geneList. If provided, the geneList will be multiplied by the weight and resorted before GSEA (default: NULL).
minGSSize	minimal size of each geneSet for analyzing
maxGSSize	maximal size of each geneSet for analyzing
eps	boundary for calculating the p value in multilevel mode
pvalueCutoff	pvalue Cutoff
pAdjustMethod	pvalue adjustment method
verbose	print message or not
nPerm	The number of permutations for the "permute" method
method	one of "sample", "permute", "multilevel"
adaptive	logical
minPerm	minimal number of permutations for the "multilevel" method
maxPerm	maximal number of permutations for the "multilevel" method
pvalThreshold	The p-value threshold for the "multilevel" method
seed	random seed for reproducibility, set to a number (or TRUE to use a fixed default seed) to make the result reproducible, or FALSE (default) to draw a random seed on each run, so results may vary between runs. The underlying permutation engine uses its own RNG seeded with this value; see enrichit::gsea() for details.
...	other parameters passed to enrichit::gsea_gson()

Value

gseaResult object

Author(s)

Yu Guangchuang

---

gson_Reactome

Description

download the latest version of Reactome and stored in a 'GSON' object

Usage

gson_Reactome(organism = "human")

Arguments

organism

one of "human", "rat", "mouse", "celegans", "yeast", "zebrafish", "fly".

Value

a 'GSON' object

Examples

## Not run: 
rec_gson <- gson_Reactome("human")

## End(Not run)

---

Network-based Reactome pathway enrichment analysis

Description

'nsePathway()' performs network-based set enrichment analysis (NSEA) on Reactome pathways. Instead of using raw gene-level scores directly, scores are propagated across the Reactome reaction graph via Random Walk with Restart (RWR), and enrichment scores are computed on the diffused ranking. Statistical significance is assessed with a whole-pipeline permutation null (gene labels of the input scores are reshuffled, the diffusion is re-run, and the enrichment scores are recomputed), so the reported p-values account for the correlation induced by network diffusion and control the false-positive rate. This captures pathway-level signals that may be diluted in individual gene-level statistics.

Usage

nsePathway(
  geneList,
  organism = "human",
  mode = c("evidence", "signed"),
  network = NULL,
  restart_prob = 0.5,
  specific_weight = FALSE,
  minGSSize = 10,
  maxGSSize = 500,
  threshold = 1e-09,
  maxIter = 100,
  nPerm = 1000,
  seed = NULL,
  verbose = TRUE,
  ...
)

Arguments

geneList	A named numeric vector of gene-level scores (e.g. log2FC). Names must be Entrez Gene IDs.
organism	Supported organism. Default is "human".
mode	Character, either "evidence" (default) or "signed". In "evidence" mode, all seeds are propagated once. In "signed" mode, positive and negative seeds are propagated separately and the difference is taken.
network	Optional pre-built network from 'prepareReactomeNetwork()'. If NULL, the network is auto-built and cached.
restart_prob	Restart probability for RWR (default: 0.5).
specific_weight	Logical, whether to apply gene specificity weighting (TF-IDF style). Default is FALSE.
minGSSize	Minimal size of each gene set for analyzing (default: 10).
maxGSSize	Maximal size of each gene set for analyzing (default: 500).
threshold	Convergence threshold for RWR (default: 1e-9).
maxIter	Maximal number of RWR iterations (default: 100).
nPerm	Number of whole-pipeline permutations used to estimate p-values and NES (default: 1000). Each permutation reshuffles the gene labels of the input scores, re-runs the RWR diffusion over the network, and recomputes the enrichment scores, giving a null distribution that accounts for the network diffusion. The smallest estimable p-value is 1/(nPerm + 1).
seed	Numeric random seed for the whole-pipeline permutation null (default: NULL uses the current R RNG state). Set a number for fully reproducible results.
verbose	Print progress messages (default: TRUE).
...	Other arguments passed to 'enrichit::nsea_gson()'.

Details

The Reactome reaction network is automatically built and cached on first call. Users can pre-build and pass a custom network via the 'network' parameter.

Value

A 'nseaResult' object (inherits from 'gseaResult').

See Also

[prepareReactomeNetwork()], [enrichit::nsea()]

Examples

## Not run: 
  data(geneList, package = "DOSE")
  nsea_res <- nsePathway(geneList)
  head(nsea_res)

## End(Not run)

---

Prepare Reactome reaction network for NSEA

Description

Builds a gene-gene interaction network from all Reactome reaction steps for a given species. Two genes are connected if they co-participate in the same reaction step. The network is returned as a column-normalized sparse matrix (dgCMatrix) suitable for Random Walk with Restart.

Usage

prepareReactomeNetwork(
  organism = "human",
  directed = FALSE,
  cores = getOption("mc.cores", 1L)
)

Arguments

organism	Supported organism: "human", "mouse", "rat", "yeast", "zebrafish", "fly", "celegans", etc. See 'graphite::pathways()' for the full list.
directed	Logical. If FALSE (default), edges are treated as undirected (reciprocal edges added).
cores	Number of CPU cores used to convert Reactome pathways to reaction edges in parallel. Defaults to 'getOption("mc.cores")', or '1' (serial) if the option is unset: the conversion is fast enough (~40 s for human) that parallelizing is optional, and serial leaves the rest of the machine untouched. The per-pathway conversions are independent and pure in-memory vectorized operations (the species identifier map is loaded once up front), so wall time scales close to linearly with 'cores'. On Windows, 'mclapply()' falls back to serial execution regardless.

Details

The network is cached via 'yulab.utils::with_cache()' so repeated calls are fast.

Value

A sparse matrix (dgCMatrix) of the Reactome reaction graph.

Examples

## Not run: 
  net <- prepareReactomeNetwork("human")

## End(Not run)

---

viewPathway

Description

view reactome pathway

Usage

viewPathway(
  pathName,
  organism = "human",
  readable = TRUE,
  foldChange = NULL,
  keyType = "ENTREZID",
  layout = "kk"
)

Arguments

pathName	pathway Name
organism	supported organism
readable	logical
foldChange	fold change
keyType	keyType of gene ID (i.e. names of foldChange, if available)
layout	graph layout

Details

plotting reactome pathway

Value

plot

Author(s)

Yu Guangchuang

---