Merge pull request #101 from chrovis/feature/chr-order

Add an ordering function for chromosome name.

src/cljam/io/bed.clj

@@ -6,7 +6,7 @@
             [proton.core :refer [as-int as-long]]
             [cljam.io.protocols :as protocols]
             [cljam.util :as util]
-            [cljam.util.chromosome :as chr-util]
+            [cljam.util.chromosome :as chr]
             [clojure.tools.logging :as logging])
   (:import [java.io BufferedReader BufferedWriter Closeable]))
 
@@ -154,7 +154,7 @@
   This function converts the coordinate into cljam style: 1-origin and inclusice-start / inclusive-end."
   [m]
   (-> m
-      (update :chr chr-util/normalize-chromosome-key)
+      (update :chr chr/normalize-chromosome-key)
       (update :start inc)
       (update-some :thick-start inc)))
 
@@ -190,9 +190,7 @@
   [xs]
   (sort-by
    (fn [m]
-     [(or (as-int (last (re-find #"(chr)?(\d+)" (:chr m)))) Integer/MAX_VALUE)
-      (or ({"X" 23 "Y" 24 "M" 25} (last (re-find #"(chr)?([X|Y|M])" (:chr m)))) Integer/MAX_VALUE)
-      (:chr m)
+     [(chr/chromosome-order-key (:chr m))
       (:start m)
       (:end m)])
    xs))

src/cljam/util/chromosome.clj

@@ -1,6 +1,7 @@
 (ns cljam.util.chromosome
   "Utilities for handling chromosome name."
-  (:require [clojure.string :as cstr]))
+  (:require [clojure.string :as cstr]
+            [proton.core :as proton]))
 
 (defn normalize-name
   [s]
@@ -48,3 +49,10 @@
   [s]
   (some? (re-matches #"^chr([0-9]{1,2}|X|Y|M|MT)"
                      (normalize-chromosome-key s))))
+
+(defn chromosome-order-key [s]
+  (if-let [[_ _ chr suffix] (re-find #"(?i)^(chr)?([1-9][0-9]*|X|Y|MT|M)(\S*)" s)]
+    (if-let [num (proton/as-int chr)]
+      [num suffix]
+      [(- Integer/MAX_VALUE (case chr "X" 4 "Y" 3 "M" 2 "MT" 1)) suffix])
+    [Integer/MAX_VALUE s]))

test/cljam/util/t_chromosome.clj

@@ -107,3 +107,26 @@
 
     "chr4_GL000257v2_alt" false
     "14_KI270723V1_random" false))
+
+(deftest chromosome-order-key
+  (are [?in ?out] (= (sort-by chr/chromosome-order-key ?in) ?out)
+    ["chr2" "chr11" "chr3" "chr1" "chr21"] ["chr1" "chr2" "chr3" "chr11" "chr21"]
+    ["2" "11" "3" "1" "21"] ["1" "2" "3" "11" "21"]
+    ["chrM_foo" "chrMT_foo"] ["chrM_foo" "chrMT_foo"]
+
+    ["chrY" "chrM_foo" "chrMT" "chrM" "chr22" "chrUn" "chrMT_foo" "chrX"]
+    ["chr22" "chrX" "chrY" "chrM" "chrM_foo" "chrMT" "chrMT_foo" "chrUn"]
+
+    ["Y" "M_foo" "MT" "M" "22" "Z" "Un" "MT_foo" "X"]
+    ["22" "X" "Y" "M" "M_foo" "MT" "MT_foo" "Un" "Z"]
+
+    ["chr19_KI270930v1_alt" "chr22_KI270879v1_alt" "chrUn_KI270425v1" "chrX" "chrUn_KI270423v1" "chrUn_KI270590v1"
+     "chr1" "chr9_GL383541v1_alt" "chr10_GL383545v1_alt" "chr6_GL000253v2_alt" "chrY" "chr9_KI270717v1_random"
+     "chr22_KI270734v1_random" "chr2" "chr18_GL383572v1_alt" "chr6_GL000251v2_alt" "chr9_KI270823v1_alt"
+     "chr19_KI270916v1_alt" "chr22_KI270875v1_alt" "chrUn_KI270330v1" "chr19_GL949753v2_alt" "chr11"
+     "chrUn_KI270312v1" "chr4_GL000008v2_random" "chr19_KI270929v1_alt"]
+    ["chr1" "chr2" "chr4_GL000008v2_random" "chr6_GL000251v2_alt" "chr6_GL000253v2_alt"
+     "chr9_GL383541v1_alt" "chr9_KI270717v1_random" "chr9_KI270823v1_alt" "chr10_GL383545v1_alt" "chr11"
+     "chr18_GL383572v1_alt" "chr19_GL949753v2_alt" "chr19_KI270916v1_alt" "chr19_KI270929v1_alt"
+     "chr19_KI270930v1_alt" "chr22_KI270734v1_random" "chr22_KI270875v1_alt" "chr22_KI270879v1_alt" "chrX" "chrY"
+     "chrUn_KI270312v1" "chrUn_KI270330v1" "chrUn_KI270423v1" "chrUn_KI270425v1" "chrUn_KI270590v1"]))