Update pairwisedistances script

inst/scripts/pairwisedistances.R

@@ -4,15 +4,17 @@ library(dplyr)
 library(castor)
 library(tibble)
 library(readr)
+library(ggplot2)
+library(purrr)
+library(tidyr)
 
-# Functions ---------------------------------------------------------------
 
+# Functions ---------------------------------------------------------------
 removeASR <- function(x) {
     x |>
         dplyr::filter(Evidence != "asr") |>
         mutate(NCBI_ID = as.character(NCBI_ID))
 }
-
 addAttributes <- function(attrDat, tipDat) {
     dplyr::left_join(
         tipDat, attrDat,
@@ -21,32 +23,149 @@ addAttributes <- function(attrDat, tipDat) {
         filter(!is.na(Attribute_value))
 }
 
+removeDups <- function(x) {
+    x |>
+        dplyr::rowwise() |>
+        dplyr::mutate(
+            sorted = paste(sort(c(node1, node2)), collapse = "_")
+        ) |>
+        dplyr::ungroup() |>
+        dplyr::filter(!duplicated(sorted)) |>
+        dplyr::select(-sorted)
+}
+getClosestTips <- function(tr) {
+    purrr::map(seq_along(tr$tip.label), function(i) {
+        if ((i %% 5000) == 0)
+            message(i)
+        res <- castor::find_nearest_tips(tr, target_tips = tr$tip.label[i])
+        tip_distances <- res[["nearest_distance_per_tip"]]
+        names(tip_distances) <- tr$tip.label
+        tip_distances <- tip_distances[-i]
+        minValue <- min(tip_distances)
+        tip_distances <- tip_distances[which(tip_distances == minValue)]
+        data.frame(
+            node2 = names(tip_distances),
+            distance = unname(tip_distances)
+        )
+    }) |>
+        purrr::set_names(tr$tip.label) |>
+        dplyr::bind_rows(.id = "node1") |>
+        removeDups()
+}
 # -------------------------------------------------------------------------
 
 ltp <- ltp()
 tr <- ltp$tree
-tip_data <- ltp$tip_data
-
+tip_data <- ltp$tip_data |>
+    group_by(taxid) |>
+    mutate(n = n()) |>
+    ungroup() |>
+    filter(n == 1) |>
+    select(-n)
 bp <- importBugphyzz()
 
-aerDat <- bp$aerophilicity |>
-    filter(Attribute_value == "aerobic") |>
+gt <- bp$`optimal ph`
+
+gt2 <- gt |>
     removeASR() |>
-    addAttributes(tipDat = tip_data)
-annotatedTips <- unique(aerDat$tip_label)
-pwd <- get_all_pairwise_distances(tree = tr, only_clades = annotatedTips)
+    select(NCBI_ID, Attribute_value)
+
+gt_tip_data <- left_join(tip_data, gt2, by = c("taxid" = "NCBI_ID")) |>
+    filter(!is.na(Attribute_value))
+
+
+
+
+# gtDat <- gt_tip_data$Attribute_value
+# names(gtDat) <- gt_tip_data$tip_label
+
+
+# discrete_trait_depth()
+#
+#
+# if (any(gtDat < 0)) {
+#     negs <- which(gtDat < 0)
+#     message("Found ", length(negs), " negatives. Dropping them.")
+#     gtDat <- gtDat[-negs]
+# }
+# gtDat <- log(gtDat + 1)
+#
+#
+#
+# tim <- system.time({
+#     res <- phylosig(tr, x = gtDat, method = "K", test = TRUE)
+# })
+
+
+
+# aerDat <- bp$aerophilicity |>
+#     removeASR() |>
+#     addAttributes(tipDat = tip_data)
+
 
-colnames(pwd) <- annotatedTips
-rownames(pwd) <- annotatedTips
+annotatedTips <- unique(gt_tip_data$tip_label)
+closeTips <- getClosestTips(tr)
 
-indices <- which(upper.tri(pwd, diag = FALSE), arr.ind = TRUE)
 
-distances <- pwd[indices]
-node1 <- rownames(pwd)[indices[, 1]] # rows
-node2 <- colnames(pwd)[indices[, 2]] # columns
+x <- select(gt_tip_data, tip_label, Attribute_value)
+
+y <- left_join(closeTips, x, by = c("node1"  = "tip_label")) |>
+    rename(node1Val = Attribute_value)
+z <- left_join(y, x, by = c("node2"  = "tip_label")) |>
+    rename(node2Val = Attribute_value)
+z <- drop_na(z)
+z$diff <- abs(z$node1Val - z$node2Val)
+
+
+z |>
+    ggplot(aes(log(distance + 1), diff)) +
+    geom_point()
+
+res <- lm(formula = distance ~ diff, data = z)
+summary(res)
+
+cor(x = z$distance, z$diff, method = "spearman")
+
+
+
+annotationsL <- split(aerDat, aerDat$Attribute_value) |>
+    map(~ pull(.x, tip_label))
+
+for (i in seq_along(annotationsL)) {
+    annName <- names(annotationsL)[i]
+
+    colName1 <- paste0(annName, "_node1")
+    colName2 <- paste0(annName, "_node2")
+
+    closeTips[[colName1]] <- closeTips$node1 %in% annotationsL[[i]]
+    closeTips[[colName2]] <- closeTips$node2 %in% annotationsL[[i]]
+
+    closeTips[[annName]] <- paste0(closeTips[[colName1]], "|", closeTips[[colName2]])
+}
+
+
+dat <- closeTips |>
+    select(-matches("_node\\d$")) |>
+    pivot_longer(
+        names_to = "Attribute", values_to = "logical",
+        cols = aerobic:last_col()
+    )
+
+
+dat2 <- dat |>
+    mutate(
+        logical = case_when(
+            grepl("FALSE", logical) ~ "FALSE",
+            TRUE ~ logical
+        )
+    ) |>
+    filter(Attribute != "aerotolerant")
+
+dat2 |>
+    count(Attribute, logical) |>
+    View()
 
-distDF <- data.frame(
-    node1 = node1,
-    node2 = node2,
-    distance = distances
-)
+dat2 |>
+    filter(logical != FALSE) |>
+    ggplot(aes(x = Attribute, y = distance)) +
+    geom_boxplot(aes(fill = logical))