fasta gap

agp_to_gff.py

@@ -45,7 +45,7 @@ def main(args):
     with open(in_agp, 'r') as fhd_in:
         with open(ou_gff, 'w') as fhd_ou:
             fhd_ou.write("##gff-version 3\n")
-            fhd_ou.write("#infile: %s\n" %in_agp)
+            fhd_ou.write("#infile: %s\n" % in_agp)
             
             for line_in in fhd_in:
                 line_in = line_in.strip()
@@ -92,6 +92,9 @@ def main(args):
                 fhd_ou.write(";".join( ["=".join([x, cols_in[x]]) for x in sorted(cols_in) if x not in gff_cols] ) )
                 fhd_ou.write("\n")
 
+
+if __name__ == '__main__':
+    main(sys.argv[1:])
         
 #1               2        3               4       5       6               7       8        9
 #SL2.50ch00      1        2191949         1       W       SL2.40sc05082   1       2191949  0
@@ -206,6 +209,3 @@ def main(args):
 #    listed using a ';' delimiter (e.g. paired-ends;align_xgenus).
 
 
-
-if __name__ == '__main__':
-    main(sys.argv[1:])
\ No newline at end of file

fasta_gap_to_gff.py

+#!/usr/bin/python
+
+import os
+import sys
+import re
+
+re_ns = re.compile('(n+)')
+
+source_name = "fasta"
+source_type = "gap"
+score, orientation, phase = [ '.', '.', '.' ]
+
+def parse_seq(ofh, seq_name, seq_seq):
+    if len(seq_seq) == 0:
+        return
+
+    print "saving chromosome", seq_name, "len", len(seq_seq)
+    hit_num = 1
+    for m in re_ns.finditer(seq_seq.lower()):
+        start_pos = m.start()
+        end_pos   = m.end()
+        match_seq = m.group()
+        diff_pos  = end_pos - start_pos
+        match_len = len(match_seq)
+        #print seq_name, start_pos, end_pos, diff_pos, match_len, match_seq
+        
+        row_id     = seq_name + '_' + str(hit_num)
+        attributes = "ID=%s;Name=%s;length=%d" % ( row_id, row_id, diff_pos )
+        
+        cols       = [ seq_name, source_name, source_type, start_pos, end_pos, score, orientation, phase, attributes ]
+        #print cols
+        
+        ofh.write("\t".join( [ str(x) for x in cols ] ) + "\n")
+        
+        hit_num += 1
+
+def main(args):
+    infasta = args[0]
+    outgff  = infasta + '.gff3'
+    
+    with open(infasta, 'r') as ifh:
+        with open(outgff, 'w') as ofh:
+            ofh.write("##gff-version 3\n")
+            ofh.write("#infile: %s\n" % infasta)
+
+            seq_name = None
+            seq_seq  = ""
+            for line in ifh:
+                line = line.strip()
+            
+                if len(line) == 0:
+                    continue
+                
+                if line[0] == ">":
+                    if seq_name is not None:
+                        parse_seq(ofh, seq_name, seq_seq)
+
+                    seq_seq  = ""
+                    seq_name = line[1:]
+                
+                else:
+                    seq_seq += line
+            
+            if seq_name is not None:
+                parse_seq(ofh, seq_name, seq_seq)
+                pass
+        
+    
+
+if __name__ == '__main__':
+    main(sys.argv[1:])
\ No newline at end of file