R/compute.tf.area.R

In phraber/lassie: Tools to select sites and sequences that represent variation in a longitudinally sampled protein alignment

#' @keywords internal

### why does this require the sequences be ordered by timepoint?
compute.tf.area <- function(tf_loss_matrix, invert=F) {

    my_tpis = as.numeric(gsub("[A-Z]", "", colnames(tf_loss_matrix), 
	ignore.case=T))

    deltaT = c(0, my_tpis[-1] - my_tpis[1:(-1+length(my_tpis))])

    tf_area = rep(NA, nrow(tf_loss_matrix))

    for (i in 1:nrow(tf_loss_matrix)) {
        if (invert) {
            tf_area[i] = sum(tf_loss_matrix[i,] * deltaT, na.rm=T)
        } else {
            tf_area[i] = sum((100-tf_loss_matrix[i,]) * deltaT, na.rm=T)
        }
    }
    ### 20180816 to do: normalize against cumulative duration
    tf_area
}