runComplexID: Annotates Hits, Performs Random Walk, and Scores Genes

In pryabinin/ComplexID: Integrated Hit Annotations and Random Walk of Complex Network

Description

Annotates hits to genes, performs random walk with restarts on a network of protein complexes, and then scores each gene in the network for its association with the phenotype of interest

Usage

runComplexID(Hits, phenoSim, promoterRange = 1e+05, eps = 1e-10,
  alpha = 0.8, upstream = 0, downstream = 0, geneBody = T,
  promoters = T, promoterTissues = "all", utr = T, eqtl = T,
  eqtlTissues = "all", enhancers = T, enhancerTissues = "all",
  loopDist = 0, non_proteins = F, geneScoring = sum, useAllTSS = T)

Arguments

Hits	Granges object with two meta data columns, or a matrix or data frame with at least 5 columns. If it is a Granges object, then the first meta data column is the site's name. The second meta data columns is a phenotype that the site is associated with. If it is a matrix or data frame, then the first column must be the Hit's name, the second column must be chromosome designation, the third column must the starting base pair position, the fourth column must be the ending base pair position (equal to starting bp position for a standard SNP) and the fifth column must a phenotype that the site is associated with. For both Grange objects and matrices/dataframes, each entry/row corresponds to one site that is associated to one phenotype. If a site is associated in multiple phenotypes then there would be multiple entries for the same site but all with different values in the phenotype column
phenoSim	matrix or data frame with two columns. The first column are names of phenotypes that match the same phenotypes found in Hits. The second column are phenotype similarity values between the phenotype in that row and the phenotype of interest (values between 0 and 1), with higher values denoting higher similarity
promoterRange	single integer greater than or equal to zero. How many bases to look upstream of a TSS of a gene in order to find a promoter region for a gene.
eps	single numeric, must be greater than zero. L1 norm threshold between current and previous interations of random walk at which to terminate the random walk
alpha	single numeric in the range of (0,1]. The weight given to the vector of initialized values for the random walk, higher value of alpha means more weight for the initialized values
upstream	single integer. By default 0. How far upstream of a transcription start site a hit can be for it to be annotated to that gene. A NULL value is equivalent to a value of zero (no upstream sites will be annotated to a gene unless they lie in a promoter region, see promoterRange parameter).
downstream	single integer. By default 0. How far downstream of a transcription start site a hit can be for it to be annotated to that gene. A NULL value is equivalent to a value of zero (no downstream sites will be annotated to a gene).
geneBody	TRUE or FALSE, by default TRUE. If TRUE, then hits will be annotated to the bodies (exons and introns) of protein coding genes. If FALSE, hits will not be annotated to those regions.
promoters	TRUE or FALSE, by default TRUE. If TRUE, then hits will be annotated to promoter regions. If FALSE, hits will not be annotated to promoter regions.
promoterTissues	character vector, by default is "all". If "all", then all promoters from all tissues will be included in the annotation, otherwise, only promoter regions from tissues specified by promoterTissues will be used for annotation.
utr	TRUE or FALSE. If TRUE then it will look for hits in the 3' and 5' UTRs of genes, otherwise it will not.
eqtl	TRUE or FALSE. By default TRUE. If TRUE, then hits may be mapped to eQTL loci, and therefore genes effected by those eQTLs be designated as associated to those hits.
eqtlTissues	character vector, by default is "all". If "all", then all eQTLs from all tissues will be included in the annotation, otherwise, only eQTL sites from tissues specified by promoterTissues will be used for annotation.
enhancers	TRUE or FALSE. By default TRUE. If TRUE, then hits may be mapped to enhancer loci and linked to genes via looping structures and promoters
enhancerTissues	character vector, by default is "all". If "all", then all enhancers from all tissues will be included in the annotation, otherwise, only enhancers regions from tissues specified by promoterTissues will be used for annotation.
loopDist	single integer. By default 0. The maximum allowable distance that an enhancer or promoter can be from a looping region to be annotated to it.
non_proteins	TRUE or FALSE. By default FALSE. If TRUE then hits may be mapped to non-protein regions, if FALSE then that annotation will not be used.
geneScoring	a function that takes a vector and outputs a single number. By default the "sum" function. This is the function that will determine the score of a gene based on the scores of the complexes that it belongs to. The input of the function is a vector of numerical values that represent the scores of the complexes that a gene belongs to. Scores are determined by the RWPCN algorithm. The output of the function should be a single numerical value.
useAllTSS	TRUE or FALSE. By default TRUE. If TRUE, then all unique transcription start sites will be considered when looking at upstream regions of a gene (for promoters and upstream regions). If FALSE, it will a single start site for a gene, namely the start of the gene.

Details

Annotates Hits to genes using a built-in annotation database. Protein coding genes, non-protein coding genes, and UTR annotations come from the ENSEMBL version 89 annotation of GRCH37. Promoter and Enhancer regions are from ENCODE annotation version 3, eQTL are from the gtexportal version 6.

After the associated genes for each endophenotype are identified, it performs a Random Walk with Restarts on a pre-constructed protein complex network as in the RWPCN method. The protein complex network was constructed in a similar way is in the RWPCN method. For a PPI we used STRING with a threshold cutoff of 700. Protein IDs in STRING were mapped to approved HUGO names using ENSEMBL and HGNC. Protein complexes were retrieved from CORUM. Any complex with no genes in the PPI was removed along with 5 of the largest complexes (more than 70 subunits)

A random walk with restarts is initialized and performed as in RWPCN then all genes in the PPI and complexes are scored according to the weights in the complex network.

Value

A list with two objects: a data frame called "scores" and a GRanges object "missingHits" The data frame "scores" has seven columns showing the scores of each gene, related to how much that gene is important to the query phenotype, as well as other information about the gene. It is ordered with the highest scoring genes first.
The first columns is the HUGO gene names, the second column are the names of the complexes that gene is part of, the third columns is the score for the gene, the fourth column says whether or not the gene was in the PPI and/or a complex, the fifth column says whether or not the gene is a protein coding gene, the sixth column are the features of that gene that have a hit in them, and the seventh column is the number of hits that were annotated to that gene. The GRanges object "missingHits" lists all of the input hits that were not mapped to any gene.

Examples

data("hits")
data("hits.pheno")
test <- runComplexID(Hits = hits,phenoSim=hits.pheno,promoterRange = 10000,upstream = 1000,downstream = 1000,utr = T)

pryabinin/ComplexID documentation built on May 8, 2019, 1:14 p.m.