R/continuous.responses.R

In antagomir/netresponse: Functional Network Analysis

# 'To invent, you need a good imagination and a pile of junk.'  -- Thomas Edison

#' @title Continuous responses
#' @description Quantify association between modes and continuous variable
#' @param annotation.vector annotation vector with discrete factor levels, and named by the samples
#' @param model NetResponse model object
#' @param method method for enrichment calculation
#' @param min.size minimum sample size for a response 
#' @param data data matrix (samples x features)
#' @return List with each element corresponding to one variable and listing the responses according to association strength
#' @author Contact: Leo Lahti \email{leo.lahti@@iki.fi}
#' @references See citation('netresponse')
#' @export
#' @keywords utilities
#' @examples res <- continuous.responses(annotation.vector = NULL, model = NULL)
continuous.responses <- function(annotation.vector, model, method = "t-test", min.size = 2,
    data = NULL) {

    if (is.null(model)) {return(NULL)}
    
    # method = 't-test'; min.size = 2; data = t(dat[, gpt])
    
    if (is.null(data) && is(model) == "NetResponseModel") {
        data <- model@datamatrix
        all.samples <- rownames(data)
    }
    
    # samples x features
    if (is.vector(data)) {
        data2 <- matrix(data)
        rownames(data2) <- names(data)
        data <- data2
        all.samples <- rownames(data)
    }
    
    if (is(model) == "NetResponseModel") {
        models <- get.subnets(model, min.size = min.size)
    } else {
        models <- model
    }
    
    associations <- enrichment.list(models, annotation.vector)
    
    associations
    
}

# associations <- continuous.responses.single(model, annotation.vector, method =
# 't.test')
continuous.responses.single <- function(model, annotation.vector, method = "t.test") {
    
    annotated.samples <- names(which(!is.na(annotation.vector)))
    annotation.data <- annotation.vector[annotated.samples]
    names(annotation.data) <- annotated.samples
    
    
    r2s <- response2sample(model)
    
    if (length(r2s) <= 1) 
        {
            return(NA)
        }  # No multimodality -> no enrichments
    
    pvals <- c()
    fold.change <- c()
    
    for (mo in seq_len(length(r2s))) {
        
        # annotated samples in the mode
        s <- intersect(r2s[[mo]], annotated.samples)
        
        # annotated samples in other modes
        sc <- intersect(unlist(r2s[-mo]), annotated.samples)
        
        if (length(na.omit(s)) > 1 && length(na.omit(sc)) > 1) {
            if (method == "t.test") {
                pval <- t.test(annotation.data[s], annotation.data[sc])$p.value
            }
            
            pvals[[mo]] <- pval
            
            fold.change[[mo]] <- mean(annotation.data[s]) -
            mean(annotation.data[sc])
        } else {
            
            warning(paste("Not enough annotated observations 
                to calculate p-values", mo))
        
            pvals[[mo]] <- NA
        
            fold.change[[mo]] <- NA
        }
    }
    
    associations <- data.frame(list(mode = seq_len(length(r2s)),
        pvalue = pvals, fold.change = fold.change))
    
    associations
    
}