From 4b9650841842bbe1798455d39bde2e6f28e80a2f Mon Sep 17 00:00:00 2001
From: Jim Procter <jprocter@issues.jalview.org>
Date: Fri, 8 Jan 2021 16:06:59 +0000
Subject: [PATCH] JAL-3794 X and N are counted as DNA characters, but we also
 check how many Ns - if there were no real nucleotide
 characters we always return protein for poly-asparagine

---
 src/jalview/util/Comparison.java |   43 +++++++++++++++++++++++++++++++-------
 1 file changed, 35 insertions(+), 8 deletions(-)

diff --git a/src/jalview/util/Comparison.java b/src/jalview/util/Comparison.java
index 286bfb2..0d945ac 100644
--- a/src/jalview/util/Comparison.java
+++ b/src/jalview/util/Comparison.java
@@ -261,7 +261,7 @@ public class Comparison
 
   /**
    * Overloaded method signature to test whether a single sequence is nucleotide
-   * (that is, more than 85% CGTA)
+   * (that is, more than 85% CGTAUNX)
    * 
    * @param seq
    * @return
@@ -274,27 +274,32 @@ public class Comparison
     }
     long ntCount = 0;
     long aaCount = 0;
+    long nCount = 0;
 
     int len = seq.getLength();
     for (int i = 0; i < len; i++)
     {
       char c = seq.getCharAt(i);
-      if (isNucleotide(c))
+      if (isNucleotide(c) || isX(c))
       {
         ntCount++;
       }
       else if (!isGap(c))
       {
         aaCount++;
+        if (isN(c))
+        {
+          nCount++;
+        }
       }
     }
     /*
      * Check for nucleotide count > 85% of total count (in a form that evades
      * int / float conversion or divide by zero).
      */
-    if (ntCount * 100 > EIGHTY_FIVE * (ntCount + aaCount))
+    if ((ntCount+nCount) * 100 > EIGHTY_FIVE * (ntCount + aaCount))
     {
-      return true;
+      return ntCount>0; // all N is considered protein. Could use a threshold here too
     }
     else
     {
@@ -317,17 +322,18 @@ public class Comparison
       return false;
     }
     // true if we have seen a nucleotide sequence
-    boolean na=false;
+    boolean na = false;
     for (SequenceI seq : seqs)
     {
       if (seq == null)
       {
         continue;
       }
-      na=true;
+      na = true;
       // TODO could possibly make an informed guess just from the first sequence
       // to save a lengthy calculation
-      if (seq.isProtein()) {
+      if (seq.isProtein())
+      {
         // if even one looks like protein, the alignment is protein
         return false;
       }
@@ -347,7 +353,6 @@ public class Comparison
     {
       c -= TO_UPPER_CASE;
     }
-
     switch (c)
     {
     case 'A':
@@ -360,6 +365,28 @@ public class Comparison
     return false;
   }
 
+  public static boolean isN(char c)
+  {
+    switch (c)
+    {
+    case 'N':
+    case 'n':
+      return true;
+    }
+    return false;
+  }
+
+  public static boolean isX(char c)
+  {
+    switch (c)
+    {
+    case 'X':
+    case 'x':
+      return true;
+    }
+    return false;
+  }
+
   /**
    * Answers true if every character in the string is one of aAcCgGtTuU, or
    * (optionally) a gap character (dot, dash, space), else false
-- 
1.7.10.2