Merge pull request #17 from gallardoalba/Add_Tsallis_metric

Add Tsallis metric

Author: esebesty

DESCRIPTION

@@ -13,7 +13,7 @@ Description: The SplicingFactory R package uses transcript-level expression
     transcript isoform diversity within samples or between conditions.
     Additionally, the package analyzes the isoform diversity data, looking for
     significant changes between conditions.
-RoxygenNote: 7.1.1
+RoxygenNote: 7.3.3
 Imports: SummarizedExperiment, methods, stats
 Suggests:
     testthat,

NAMESPACE

@@ -6,6 +6,7 @@ export(calculate_entropy)
 export(calculate_gini)
 export(calculate_inverse_simpson)
 export(calculate_simpson)
+export(calculate_tsallis_entropy)
 import(methods)
 import(stats)
 importFrom(SummarizedExperiment,SummarizedExperiment)

NEWS.md

@@ -1,3 +1,9 @@
+# SplicingFactory 1.3.2 (dev)
+
+* Added Tsallis entropy as a diversity metric, with full documentation and examples.
+* Users can now set the `q` parameter for Tsallis entropy in `calculate_diversity()` and `calculate_method()`.
+* Documentation and vignette updated to reflect this new feature.
+
 # SplicingFactory 1.3.1 (dev)
 
 * Citation update

R/calculate_diversity.R

@@ -6,8 +6,8 @@
 #'   input dataset with transcript-level expression values. The values in
 #'   \code{x} are grouped into genes based on this vector.
 #' @param method Method to use for splicing diversity calculation, including
-#'   naive entropy (\code{naive}), Laplace entropy (\code{laplace}), Gini index
-#'   (\code{gini}), Simpson index (\code{simpson}) and inverse Simpson index
+#'   naive entropy (\code{naive}), Laplace entropy (\code{laplace}), Tsallis entropy (\code{tsallis}),
+#'   Gini index (\code{gini}), Simpson index (\code{simpson}) and inverse Simpson index
 #'   (\code{invsimpson}). The default method is Laplace entropy.
 #' @param norm If \code{TRUE}, the entropy values are normalized to the number
 #'   of transcripts for each gene. The normalized entropy values are always
@@ -21,6 +21,11 @@
 #'    to use for diversity calculations.
 #' @param verbose If \code{TRUE}, the function will print additional diagnostic
 #'    messages, besides the warnings and errors.
+#' @param q Tsallis entropy parameter (q ≥ 0). Only used if method = "tsallis".
+#'   Default is 2. Must be a single scalar value.
+#'   Tsallis entropy is a generalization that encompasses multiple diversity measures:
+#'   q = 0 gives species richness, q = 1 gives Shannon entropy, and other q values
+#'   give related diversity indices (e.g., Simpson index at q=2).
 #' @return Gene-level splicing diversity values in a \code{SummarizedExperiment}
 #'    object.
 #' @import methods
@@ -35,7 +40,7 @@
 #' diversity values for each gene in each sample. These diversity values can be
 #' used to investigate the dominance of a specific transcript for a gene,
 #' the diversity of transcripts in a gene, and analyze changes in diversity.
-#'
+#' 
 #' There are a number of diversity values implemented in the package. These
 #' include the following:
 #' \itemize{
@@ -44,6 +49,9 @@
 #'     values mean a more diverse set of transcripts for a gene.
 #'   \item Laplace entropy: Shannon entropy where the transcript frequencies are
 #'     replaced by a Bayesian estimate, using Laplace's prior.
+#'   \item Tsallis entropy: A generalization of Shannon entropy, parameterized by q (q ≥ 0). 
+#'     q = 0 gives species richness, q → 1 gives Shannon entropy, q ≠ 1 gives Tsallis entropy.
+#'     The default q is 2.
 #'   \item Gini index: a measure of statistical dispersion originally used in
 #'     economy. This measurement ranges from 0 (complete equality) to 1
 #'     (complete inequality). A value of 1 (complete inequality) means a single
@@ -73,7 +81,7 @@
 #' # calculating normalized Laplace entropy
 #' result <- calculate_diversity(x, gene, method = "laplace", norm = TRUE)
 calculate_diversity <- function(x, genes = NULL, method = "laplace", norm = TRUE,
-                                tpm = FALSE, assayno = 1, verbose = FALSE) {
+                                tpm = FALSE, assayno = 1, verbose = FALSE, q = 2) {
   if (!(is.matrix(x) || is.data.frame(x) || is.list(x) || is(x, "DGEList") ||
     is(x, "RangedSummarizedExperiment") || is(x, "SummarizedExperiment"))) {
     stop("Input data type is not supported! Please use `?calculate_diversity`
@@ -143,7 +151,7 @@ calculate_diversity <- function(x, genes = NULL, method = "laplace", norm = TRUE
     stop("The number of rows is not equal to the given gene set.", call. = FALSE)
   }
 
-  if (!(method %in% c("naive", "laplace", "gini", "simpson", "invsimpson"))) {
+  if (!(method %in% c("naive", "laplace", "tsallis", "gini", "simpson", "invsimpson"))) {
     stop("Invalid method. Please use `?calculate_diversity` to see the possible
          arguments and details.",
       call. = FALSE
@@ -168,18 +176,28 @@ calculate_diversity <- function(x, genes = NULL, method = "laplace", norm = TRUE
             have any effect on the calculation.", call. = FALSE)
   }
 
-  result <- calculate_method(x, genes, method, norm, verbose = verbose)
+  result <- calculate_method(x, genes, method, norm, verbose = verbose, q = q)
 
+  # Prepare assay and row/col data
   result_assay <- result[, -1, drop = FALSE]
-  rownames(result_assay) <- result[, 1]
   result_rowData <- data.frame(genes = result[, 1], row.names = result[, 1])
-  result_colData <- data.frame(samples = colnames(x), row.names = colnames(x))
+
+  # For Tsallis with scalar q, columns correspond to samples only
+  col_ids <- colnames(x)
+  row_ids <- as.character(result[, 1])
+  result_colData <- data.frame(samples = col_ids, row.names = col_ids)
+  colnames(result_assay) <- col_ids
+  rownames(result_assay) <- row_ids
+
   result_metadata <- list(method = method, norm = norm)
+  if (method == "tsallis") result_metadata$q <- q
 
-  result <- SummarizedExperiment(assays = list(diversity = result_assay),
-                                 rowData = result_rowData,
-                                 colData = result_colData,
-                                 metadata = result_metadata)
+  result <- SummarizedExperiment(
+    assays = list(diversity = result_assay),
+    rowData = result_rowData,
+    colData = result_colData,
+    metadata = result_metadata
+  )
 
   return(result)
 }

R/calculate_method.R

@@ -6,13 +6,18 @@
 #'   input dataset with transcript-level expression values. The values in
 #'   \code{x} are grouped into genes based on this vector.
 #' @param method Method to use for splicing diversity calculation, including
-#'   naive entropy (\code{naive}), Laplace entropy (\code{laplace}), Gini index
-#'   (\code{gini}), Simpson index (\code{simpson}) and inverse Simpson index
+#'   naive entropy (\code{naive}), Laplace entropy (\code{laplace}), Tsallis entropy (\code{tsallis}),
+#'   Gini index (\code{gini}), Simpson index (\code{simpson}) and inverse Simpson index
 #'   (\code{invsimpson}). The default method is Laplace entropy.
 #' @param norm If \code{TRUE}, the entropy values are normalized to the number
 #'   of transcripts for each gene. The normalized entropy values are always
 #'   between 0 and 1. If \code{FALSE}, genes cannot be compared to each other,
 #'   due to possibly different maximum entropy values.
+#' @param q Tsallis entropy parameter (q ≥ 0). Only used if method = "tsallis".
+#'   Default is 2. Must be a single scalar value.
+#'   Tsallis entropy is a generalization that encompasses multiple diversity measures:
+#'   q = 0 gives species richness, q = 1 gives Shannon entropy, and other q values
+#'   give related diversity indices (e.g., Simpson index at q=2).
 #' @param verbose If \code{TRUE}, the function will print additional diagnostic
 #'   messages, besides the warnings and errors.
 #' @return Gene-level splicing diversity values in a \code{data.frame}, where
@@ -23,7 +28,8 @@
 #' transcript-level expression values, aggregated by the genes defined in the
 #' \code{genes} parameter.
 #' @import stats
-calculate_method <- function(x, genes, method, norm = TRUE, verbose = FALSE) {
+calculate_method <- function(x, genes, method, norm = TRUE, verbose = FALSE, q = 2) {
+
     if (method == "naive") {
         x <- aggregate(x, by = list(genes), calculate_entropy, norm = norm)
     }
@@ -33,6 +39,31 @@ calculate_method <- function(x, genes, method, norm = TRUE, verbose = FALSE) {
                        pseudocount = 1)
     }
 
+    if (method == "tsallis") {
+        # Note: q must be a scalar value (required for statistical testing)
+        # calculate_tsallis_entropy enforces length(q) == 1
+        gene_levels <- unique(genes)
+        coln <- colnames(x)
+        rown <- gene_levels
+        tsallis_row <- function(gene) {
+            idx <- which(genes == gene)
+            sapply(seq_len(ncol(x)), function(j) {
+                calculate_tsallis_entropy(x[idx, j], q = q, norm = norm)
+            })
+        }
+        result_mat <- t(vapply(gene_levels, tsallis_row, FUN.VALUE = numeric(ncol(x))))
+        colnames(result_mat) <- coln
+        rownames(result_mat) <- rown
+        out_df <- data.frame(Gene = rown, result_mat, check.names = FALSE)
+        if (all(rowSums(!is.na(result_mat)) == 0)) {
+            out_df <- data.frame(Gene=character(0))
+            for (nm in coln) out_df[[nm]] <- numeric(0)
+            x <- out_df
+            return(x)
+        }
+        x <- out_df
+    }
+
     if (method == "gini") {
         x <- aggregate(x, by = list(genes), calculate_gini)
     }