Molecular phenotypes from single and double knockdowns in RNAi screen

Description

Single and double siRNA knockdowns were performed for genes and gene pairs. Multiple molecular phenotypes, such as the number of cells, cell size, nucleus size, etc., were measured.

Value

A data matrix with each row a knockdown experiment.

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

Examples

## Not run: 
library (systemfit)
library (qvalue)

data (RNAi)
data (tested_pairs) # gene pairs tested in the RNAi knockdown assay

# extract gene names and put in a vector
genelist <- union(unique(RNAi$template_gene),unique(RNAi$query_gene))
genelist <- genelist[!((genelist=="empty")|(genelist=="NT"))]

# create the interaction terms for linear model
sorted_tested_pairs <- apply(tested_pairs,1,
	function(x){if (x[1]>x[2]) return (c(x[2],x[1])) 
	else return(c(x[1],x[2]))})
pairs_names <- apply(sorted_tested_pairs,2,
	function(x) {paste(x[1],x[2],sep=":")})

# create vector of covariates
# using batch3 as baseline
regressors <- c("batch1","batch2","batch4",genelist,pairs_names)

# construct the design matrix
my_matrix=constructDesignMatrix(data=RNAi, covariates=regressors)

# n (cell number) and csize (cell size) are on log2 scale already
# need to transform nsize (nucleus size) to original scale
RNAi.tmp <- RNAi
RNAi$nsize <- 2^RNAi.tmp$nsize
rm (RNAi.tmp)

# create formula from column names
# using all columns
#eqlog2n <- as.formula (paste ("RNAi$n ~ ", 
#	paste (colnames (my_matrix), collapse="+"), sep=''))
#eqlog2csize <- as.formula (paste ("RNAi$csize ~ ", 
#	paste (colnames (my_matrix), collapse="+"), sep=''))
#eqnsize <- as.formula (paste ("RNAi$nsize ~ ", 
#	paste (colnames (my_matrix), collapse="+"), sep=''))
	
# test run with the first 500 columns
eqlog2n <- as.formula (paste ("RNAi$n ~ ", 
	paste (colnames (my_matrix)[1:500], collapse="+"), sep=''))
eqlog2csize <- as.formula (paste ("RNAi$csize ~ ", 
	paste (colnames (my_matrix)[1:500], collapse="+"), sep=''))
eqnsize <- as.formula (paste ("RNAi$nsize ~ ", 
	paste (colnames (my_matrix)[1:500], collapse="+"), sep=''))
	
system <- list (cell.number = eqlog2n, cell.size = eqlog2csize, nuc.size=eqnsize)

# perform seemingly unrelated regression
fitsur <- systemfit (system, "SUR", data=cbind (RNAi, my_matrix), maxit=100)

# extract coefficient estimates
log2n_fitsur_coef <- coef (summary (fitsur$eq[[1]]))
log2csize_fitsur_coef <- coef (summary (fitsur$eq[[2]]))
nsize_fitsur_coef <- coef (summary (fitsur$eq[[3]]))

# compute q values
log2n_coef_q <- qvalue (log2n_fitsur_coef[,4])$qvalues
log2csize_coef_q <- qvalue (log2csize_fitsur_coef[,4])$qvalues
nsize_coef_q <- qvalue (nsize_fitsur_coef[,4])$qvalues

# build three matrices of results
log2n_fitsur_coef <- data.frame (log2n_fitsur_coef, qvalue=log2n_coef_q)
colnames (log2n_fitsur_coef) <- c("Estimate", "StdError", "tValue", "pValue", "qValue")
dim (log2n_fitsur_coef)
head (log2n_fitsur_coef)

log2csize_fitsur_coef <- data.frame (log2csize_fitsur_coef, qvalue=log2csize_coef_q)
colnames (log2csize_fitsur_coef) <- c("Estimate", "StdError", "tValue", "pValue", "qValue")
dim (log2csize_fitsur_coef)
head (log2csize_fitsur_coef)

nsize_fitsur_coef <- data.frame (nsize_fitsur_coef, qvalue=nsize_coef_q)
colnames (nsize_fitsur_coef) <- c("Estimate", "StdError", "tValue", "pValue", "qValue")
dim (nsize_fitsur_coef)
head (nsize_fitsur_coef)

## End(Not run)

Compute smallworldness of a graph

Description

This function computes the smallworldness of a graph.

Usage

computeSmallWorldness(g, n, m, nrep = 1000)

Arguments

Details

For a graph g with n nodes and m edges, the smallworldness S is defined as in Humphries and Gurney (2008):

S = (C_g / C_{rand}) / (L_g / L_{rand}),

where C_g and C_{rand} are the clustering coefficient of g and that of a random graph with the same number of nodes and edges as g, respectively. Also, L_g and L_{rand} are the mean shortest path length of g and that of the same random graph, respectively.

Here, in order to estimate C_{rand} and L_{rand}, this function generates a large number of random graphs with n nodes and m edges under the Erdos-Renyi model (Erdos and Renyi, 1959), such that each edge is created with the same probability as the nodes in g. This function then computes C and L for each random graph, and takes the average as the estimate for C_{rand} and L_{rand}.

Value

A scalar of smallworldness.

Author(s)

Audrey Q. Fu

References

Humphries, M. D. and Gurney, K. Network 'small-world-ness': a quantitative method for determining canonical network equivalence. PLoS ONE 3, e0002051 (2008).

Erdos, P. and Renyi, A. On random graphs. Publ. Math. 6, 290-297 (1959).

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

Examples

library (igraph)
# compute smallworldness for the design graph
data (tested_pairs)
# build the graph object
g <- graph.edgelist (as.matrix (tested_pairs), directed=FALSE)
summary (g)  # 67 nodes and 1508 edges
# compute smallworldness
computeSmallWorldness (g, n=67, m=1508)

Survival analysis for pairs of genes

Description

This function counts the number of individuals with different mutation patterns, estimates the median survival time for each mutation pattern, and computes the p values.

Usage

computeSurvivalPValueForGenePairSet.output(file.out, 
	gene.pairs, data.mut, data.surv, 
	colTime = 2, colStatus = 3, 
	type.gene1 = (-1), type.gene2 = (-1), 
	groups = c("All", "Two"), 
	PRINT = FALSE, PRINT.INDEX = FALSE)

Arguments

Value

Data frame containing the following columns (if groups="Two"):

Author(s)

Audrey Q. Fu, Xiaoyue Wang

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

Examples

## Not run: 
data (mutations)
data (survival)

# compute p values for gene pairs tested in the RNAi knockdown assay
data (tested_pairs)

# compute p values for the gain & loss combination
# and compare only cases of single mutations with cases of double mutations;
# results are written to file tmp.txt under current directory
computeSurvivalPValueForGenePairSet.output (file.out="tmp.txt", 
	tested_pairs, data.mut=mutations, data.surv=survival, 
	type.gene1=1, type.gene2=(-1), groups="Two")


## End(Not run)

Survival analysis for pairs of genes (with matched individuals)

Description

This function is similar to computeSurvivalPValueForGenePairSet.output, except that individuals in data.mut and data.surv should match, and that gene.pairs contains four columns: gene1, mutation type of gene1, gene2, mutation type of gene2.

Usage

computeSurvivalPValueGenePairAll.output(file.out, 
	gene.pairs, data.mut, data.surv, 
	colTime = 2, colStatus = 3, 
	groups = c("All", "Two"), 
	PRINT = FALSE, PRINT.INDEX = FALSE)

Arguments

Author(s)

Audrey Q. Fu, Xiaoyue Wang

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

See Also

Called by computeSurvivalPValueForGenePairSet.output

Survival analysis for one pair of genes

Description

This function performs survival analysis, similar to function computeSurvivalPValueForGenePairSet.output, but for one pair of genes.

Usage

computeSurvivalPValueOneGenePair(data.mut, data.surv, 
	colTime = 2, colStatus = 3, 
	type.gene1 = (-1), type.gene2 = (-1), 
	groups = c("All", "Two"), 
	compare = c("Both", "Gene1", "Gene2"), 
	PLOT = FALSE, PRINT = FALSE, 
	pvalue.text.x = 10, pvalue.text.y = 0.1, 
	legend.x = 150, legend.y = 1)

Arguments

Value

The output contains the same info as described in computeSurvivalPValueForGenePairSet.output.

Author(s)

Audrey Q. Fu, Xiaoyue Wang

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

See Also

computeSurvivalPValueForGenePairSet.output

Write results from survival analysis to output for one pair of genes

Description

This function is similar to computeSurvivalPValueOneGenePair, except that it writes the analysis results directly to output file and does not allow for plotting the survival curves.

Usage

computeSurvivalPValueOneGenePair.output(file.out, genes.info, 
	data.mut, data.surv, colTime = 2, colStatus = 3, 
	groups = c("All", "Two"), PRINT = FALSE)

Arguments

Value

A vector of values from the survival analysis, as described in computeSurvivalPValueForGenePairSet.output

Author(s)

Audrey Q. Fu, Xiaoyue Wang

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

See Also

computeSurvivalPValueForGenePairSet.output

Generate a design matrix from raw RNAi data.

Description

This function takes the raw RNAi data as input and generates a design matrix for regression. Specifically written for the format of the data set RNAi, which contains four batches. This R function will use batch3 as the baseline.

Usage

constructDesignMatrix(data, covariates)

Arguments

Value

A design matrix. The number of rows is the same as that of the data set RNAi, and the number of columns is the same as the length of covariates.

Examples

## See example in documentation for the data set RNAi.

Genetic mutation data in patients.

Description

Data frame that contains mutation patterns in multiple genes across multiple patients.

Format

A data frame with 85 rows and 951 columns. Each row is a gene. The first column contains gene names, and each of the other columns contains the mutation pattern in an individual: 0 for no mutation, 1 amplification and -1 deletion.

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828.

Examples

data(mutations)

Find matched individuals in mutation and survival data

Description

This functions finds matched individuals in data.mut and data.surv, and outputs the two data sets with only matched individuals.

Usage

processDataMutSurv(data.mut, data.surv, colTime = 2, colStatus = 3)

Arguments

Value

A list of two data frames, data.mut and data.surv. Format of the data frames is the same as input, except that the individuals in the two data frames are matched.

Author(s)

Audrey Q. Fu, Xiaoyue Wang

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

See Also

computeSurvivalPValueForGenePairSet.output

Patient survival data.

Description

Data set that contains the survival time (in months), survival status and other information of patients.

Format

A data frame with 950 observations on the following 5 variables.

CaseID

A vector of character strings

OverallSurvivalMonths

A numeric vector

OverallSurvivalStatus

A factor with levels DECEASED LIVING

MutationCount

A numeric vector

FractionOfCopyNumberAlteredGenome

A numeric vector

Source

Data were downloaded from http://www.cbioportal.org/.

References

Data were described and analyzed in Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828.

Examples

data(survival)

Compute the p and q values of all pairwise gene mutation patterns

Description

This function computes the p and q values of all pairwise gene mutation patterns. Patterns include both genes losing their function, one gene gaining function and the other losing function, both genes gaining function, and the two genes being mutually exclusive.

Usage

testMutationalPatternAll.wrapper(data, QVALUE = TRUE, PRINT = FALSE)

Arguments

Value

A list of two matrices, one containing the p values, and the other the q values (if the QVALUE argument set to TRUE). Each matrix has the following columns: gene 1, gene 2, p (or q) value of the loss & loss, gain & loss, loss & gain, gain & gain, and mutually exclusive combination.

Author(s)

Audrey Fu, Xiaoyue Wang

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

Examples

data (mutations)
mut.pqvalues <- testMutationalPatternAll.wrapper (data=mutations, QVALUE=TRUE)
summary (mut.pqvalues)
dim (mut.pqvalues$pvalues)
dim (mut.pqvalues$qvalues)
mut.pqvalues$pvalues[1:10,]

Gene pairs tested in the double knockdown assay.

Description

It contains two columns of gene names.

Format

A data frame with 1508 observations on the following 2 variables.

V1

a character vector

V2

a character vector

References

Wang, X., Fu, A. Q., McNerney, M. and White, K. P. (2014). Widespread genetic epistasis among breast cancer genes. Nature Communications. 5 4828. doi: 10.1038/ncomms5828

Examples

data(tested_pairs)
## see documentation for dataset \code{\link{RNAi}}