Merge pull request #35 from bewt85/bt5_496468_qm_bases

496468: Allow '?' in fasta input

src/alignment-file.c

@@ -181,6 +181,18 @@ int build_reference_sequence_and_truncate(char reference_sequence[], char filena
   return 1;
 }
 
+int is_unknown(char base)
+{
+  switch (toupper(base)) {
+    case 'N':
+    case '-':
+    case '?':
+      return 1;
+    default:
+      return 0;
+  }
+}
+
 int detect_snps(char reference_sequence[], char filename[], size_t length_of_genome)
 {
   int i;
@@ -199,12 +211,12 @@ int detect_snps(char reference_sequence[], char filename[], size_t length_of_gen
     for(i = 0; i < length_of_genome; i++)
       {
 	// If there is an indel in the reference sequence, replace with the first proper base you find
-	if((reference_sequence[i] == '-' && seq->seq.s[i] != '-' ) || (toupper(reference_sequence[i]) == 'N' && seq->seq.s[i] != 'N' ))
+	if(is_unknown(reference_sequence[i]) && !is_unknown(seq->seq.s[i]))
 	  {
 	    reference_sequence[i] = toupper(seq->seq.s[i]);
 	  }
 
-	if(reference_sequence[i] != '*' && seq->seq.s[i] != '-' && toupper(seq->seq.s[i]) != 'N' && reference_sequence[i] != toupper(seq->seq.s[i]))
+	if(! is_unknown(reference_sequence[i]) && reference_sequence[i] != '*' && ! is_unknown(seq->seq.s[i]) && (reference_sequence[i] != toupper(seq->seq.s[i])))
 	  {
 	    reference_sequence[i] = '*';
 	    number_of_snps++;

src/alignment-file.h

@@ -22,6 +22,7 @@
 
 #include "kseq.h"
 
+int is_unknown(char base);
 int detect_snps(char reference_sequence[],  char filename[], size_t length_of_genome);
 int line_length(FILE * alignment_file_pointer);
 int build_reference_sequence(char reference_sequence[], char filename[]);

src/parse-phylip.c

@@ -42,7 +42,7 @@ int does_column_contain_snps(int snp_column, char reference_base)
 			return 0;	
 		}
 
-		if(sequences[i][snp_column]  != '-' && toupper(sequences[i][snp_column])  != 'N' && sequences[i][snp_column] != reference_base)
+		if(!is_unknown(sequences[i][snp_column]) && sequences[i][snp_column] != reference_base)
 		{
 			return 1;
 		}

src/vcf.c

@@ -121,7 +121,7 @@ char * alternative_bases(char reference_base, char * bases_for_snp, int number_o
 	char * alt_bases = calloc(MAXIMUM_NUMBER_OF_ALT_BASES+1, sizeof(char));
 	for(i=0; i< number_of_samples; i++ )
 	{
-		if((bases_for_snp[i] != reference_base) && (bases_for_snp[i] != '-') && (toupper(bases_for_snp[i]) != 'N') )
+		if(!is_unknown(bases_for_snp[i]) && (bases_for_snp[i] != reference_base))
 		{
 			if(check_if_char_in_string(alt_bases, bases_for_snp[i], num_alt_bases) == 0)
 			{
@@ -145,7 +145,7 @@ char * format_allele_index(char base, char reference_base, char * alt_bases)
 	assert(length_of_alt_bases < 100);
 	char * result = calloc(3, sizeof(char));
 	int index;
-	if (reference_base == base || toupper(base) == 'N' || base == '-')
+	if (reference_base == base || is_unknown(base))
 	{
 		sprintf(result, "0");
 	}