R/preprocessing.r

In priesgo/TCGAome: MultiOmics (just RNAseq and RPPA...) Data Analyzer on TCGA data

### Data preprocessing
preprocess_data <- function(X, Y, correlation.thr = 0.7) {
    futile.logger::flog.info("Starts preprocessing...")

    # Prepares matrices for mixOmics
    X <- as.data.frame(X)
    Y <- as.data.frame(Y)

    futile.logger::flog.info("No. of RNAseq variables before preprocessing: %s", dim(X)[2])
    futile.logger::flog.info("No. of RPPA variables before preprocessing: %s", dim(Y)[2])

    # Scales matrices
    X <- scale(X)
    Y <- scale(Y)

    # Remove variables with zero-variance (20169 genes and 142 proteins) variance.thr = 0.01
    X <- X[, apply(X, 2, function(x) 0 != var(if (is.factor(x))
        as.integer(x) else x)), drop = F]
    Y <- Y[, apply(Y, 2, function(x) 0 != var(if (is.factor(x))
        as.integer(x) else x)), drop = F]
    X <- X[, - mixOmics::nearZeroVar(X)$Position]
    # Y = Y[, -nearZeroVar(Y)$Position] # there are no nearzero variables in Y

    futile.logger::flog.info("No. of RNAseq variables after removing zero and near zero variance variables: %s", dim(X)[2])
    futile.logger::flog.info("No. of RPPA variables after removing zero and near zero variance variables: %s", dim(Y)[2])

    # Remove variables having any missing value TODO: evaluate other approaches to deal with missing values
    X <- X[, apply(X, 2, function(x) !any(is.na(x))), drop = F]
    Y <- Y[, apply(Y, 2, function(x) !any(is.na(x))), drop = F]

    futile.logger::flog.info("No. of RNAseq variables after removing variables with missing values: %s", dim(X)[2])
    futile.logger::flog.info("No. of RPPA variables after removing variables with missing values: %s", dim(Y)[2])

    # Removes correlated variables using a threshold of 0.7 (14179 genes and 121 proteins) correlation.thr = 0.7 ## This value is in the literature from Le Cao
    X.cor <- cor(X)
    X.cor[!lower.tri(X.cor)] <- 0  # sets upper triangle and diagonal to 0
    X <- X[, !apply(X.cor, 2, function(x) any(x > correlation.thr))]
    Y.cor <- cor(Y)
    Y.cor[!lower.tri(Y.cor)] <- 0  # sets upper triangle and diagonal to 0
    Y <- Y[, !apply(Y.cor, 2, function(x) any(x > correlation.thr))]

    futile.logger::flog.info("Correlation threshold: %s", correlation.thr)
    futile.logger::flog.info("No. of RNAseq variables after removing duplicate variables with high correlation: %s", dim(X)[2])
    futile.logger::flog.info("No. of RPPA variables after removing duplicate variables with high correlation: %s", dim(Y)[2])

    futile.logger::flog.info("Preprocessing finished.")

    # returns preprocessed matrices X, Y
    list(X = X, Y = Y)
}