datamodel/compbio/data/sequence/SequenceUtil.java

   1 /*\r
   2  * @(#)SequenceUtil.java 1.0 September 2009 Copyright (c) 2009 Peter Troshin\r
   3  * Jalview Web Services version: 2.0 This library is free software; you can\r
   4  * redistribute it and/or modify it under the terms of the Apache License\r
   5  * version 2 as published by the Apache Software Foundation This library is\r
   6  * distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY;\r
   7  * without even the implied warranty of MERCHANTABILITY or FITNESS FOR A\r
   8  * PARTICULAR PURPOSE. See the Apache License for more details. A copy of the\r
   9  * license is in apache_license.txt. It is also available here: see:\r
  10  * http://www.apache.org/licenses/LICENSE-2.0.txt Any republication or derived\r
  11  * work distributed in source code form must include this copyright and license\r
  12  * notice.\r
  13  */\r
  14 \r
  15 package compbio.data.sequence;\r
  16 \r
  17 import java.io.BufferedReader;\r
  18 import java.io.BufferedWriter;\r
  19 import java.io.Closeable;\r
  20 import java.io.File;\r
  21 import java.io.FileInputStream;\r
  22 import java.io.IOException;\r
  23 import java.io.InputStream;\r
  24 import java.io.InputStreamReader;\r
  25 import java.io.OutputStream;\r
  26 import java.io.OutputStreamWriter;\r
  27 import java.util.ArrayList;\r
  28 import java.util.HashMap;\r
  29 import java.util.HashSet;\r
  30 import java.util.List;\r
  31 import java.util.Map;\r
  32 import java.util.Scanner;\r
  33 import java.util.Set;\r
  34 import java.util.SortedSet;\r
  35 import java.util.TreeSet;\r
  36 import java.util.logging.Level;\r
  37 import java.util.regex.Matcher;\r
  38 import java.util.regex.Pattern;\r
  39 \r
  40 import compbio.util.Util;\r
  41 \r
  42 /**\r
  43  * Utility class for operations on sequences\r
  44  * \r
  45  * @author Petr Troshin\r
  46  * @version 1.0\r
  47  */\r
  48 public final class SequenceUtil {\r
  49 \r
  50         /**\r
  51          * A whitespace character: [\t\n\x0B\f\r]\r
  52          */\r
  53         public static final Pattern WHITE_SPACE = Pattern.compile("\\s");\r
  54 \r
  55         /**\r
  56          * A digit\r
  57          */\r
  58         public static final Pattern DIGIT = Pattern.compile("\\d");\r
  59 \r
  60         /**\r
  61          * Non word\r
  62          */\r
  63         public static final Pattern NONWORD = Pattern.compile("\\W");\r
  64 \r
  65         /**\r
  66          * Valid Amino acids\r
  67          */\r
  68         public static final Pattern AA = Pattern.compile("[ARNDCQEGHILKMFPSTWYV]+",\r
  69                         Pattern.CASE_INSENSITIVE);\r
  70 \r
  71         /**\r
  72          * inversion of AA pattern\r
  73          */\r
  74         public static final Pattern NON_AA = Pattern.compile(\r
  75                         "[^ARNDCQEGHILKMFPSTWYV]+", Pattern.CASE_INSENSITIVE);\r
  76 \r
  77         /**\r
  78          * Same as AA pattern but with two additional letters - XU\r
  79          */\r
  80         public static final Pattern AMBIGUOUS_AA = Pattern.compile(\r
  81                         "[ARNDCQEGHILKMFPSTWYVXU]+", Pattern.CASE_INSENSITIVE);\r
  82 \r
  83         /**\r
  84          * Nucleotides a, t, g, c, u\r
  85          */\r
  86         public static final Pattern NUCLEOTIDE = Pattern.compile("[AGTCU]+",\r
  87                         Pattern.CASE_INSENSITIVE);\r
  88 \r
  89         /**\r
  90          * Ambiguous nucleotide\r
  91          */\r
  92         public static final Pattern AMBIGUOUS_NUCLEOTIDE = Pattern.compile(\r
  93                         "[AGTCRYMKSWHBVDNU]+", Pattern.CASE_INSENSITIVE); // see IUPAC\r
  94         /**\r
  95          * Non nucleotide\r
  96          */\r
  97         public static final Pattern NON_NUCLEOTIDE = Pattern.compile("[^AGTCU]+",\r
  98                         Pattern.CASE_INSENSITIVE);\r
  99 \r
 100         private SequenceUtil() {\r
 101         } // utility class, no instantiation\r
 102 \r
 103         /*\r
 104          * public static void write_PirSeq(OutputStream os, FastaSequence seq)\r
 105          * throws IOException { BufferedWriter pir_out = new BufferedWriter(new\r
 106          * OutputStreamWriter(os)); pir_out.write(">P1;" + seq.getId() +\r
 107          * SysPrefs.newlinechar); pir_out.write(seq.getSequence() +\r
 108          * SysPrefs.newlinechar); pir_out.close(); } public static void\r
 109          * write_FastaSeq(OutputStream os, FastaSequence seq) throws IOException {\r
 110          * BufferedWriter fasta_out = new BufferedWriter( new\r
 111          * OutputStreamWriter(os)); fasta_out.write(">" + seq.getId() +\r
 112          * SysPrefs.newlinechar); fasta_out.write(seq.getSequence() +\r
 113          * SysPrefs.newlinechar); fasta_out.close(); }\r
 114          */\r
 115 \r
 116         /**\r
 117          * @return true is the sequence contains only letters a,c, t, g, u\r
 118          */\r
 119         public static boolean isNucleotideSequence(final FastaSequence s) {\r
 120                 return SequenceUtil.isNonAmbNucleotideSequence(s.getSequence());\r
 121         }\r
 122 \r
 123         /**\r
 124          * Ambiguous DNA chars : AGTCRYMKSWHBVDN // differs from protein in only one\r
 125          * (!) - B char\r
 126          */\r
 127         public static boolean isNonAmbNucleotideSequence(String sequence) {\r
 128                 sequence = SequenceUtil.cleanSequence(sequence);\r
 129                 if (SequenceUtil.DIGIT.matcher(sequence).find()) {\r
 130                         return false;\r
 131                 }\r
 132                 if (SequenceUtil.NON_NUCLEOTIDE.matcher(sequence).find()) {\r
 133                         return false;\r
 134                         /*\r
 135                          * System.out.format("I found the text starting at " +\r
 136                          * "index %d and ending at index %d.%n", nonDNAmatcher .start(),\r
 137                          * nonDNAmatcher.end());\r
 138                          */\r
 139                 }\r
 140                 final Matcher DNAmatcher = SequenceUtil.NUCLEOTIDE.matcher(sequence);\r
 141                 return DNAmatcher.find();\r
 142         }\r
 143 \r
 144         /**\r
 145          * Removes all whitespace chars in the sequence string\r
 146          * \r
 147          * @param sequence\r
 148          * @return cleaned up sequence\r
 149          */\r
 150         public static String cleanSequence(String sequence) {\r
 151                 assert sequence != null;\r
 152                 final Matcher m = SequenceUtil.WHITE_SPACE.matcher(sequence);\r
 153                 sequence = m.replaceAll("").toUpperCase();\r
 154                 return sequence;\r
 155         }\r
 156 \r
 157         /**\r
 158          * Removes all special characters and digits as well as whitespace chars\r
 159          * from the sequence\r
 160          * \r
 161          * @param sequence\r
 162          * @return cleaned up sequence\r
 163          */\r
 164         public static String deepCleanSequence(String sequence) {\r
 165                 sequence = SequenceUtil.cleanSequence(sequence);\r
 166                 sequence = SequenceUtil.DIGIT.matcher(sequence).replaceAll("");\r
 167                 sequence = SequenceUtil.NONWORD.matcher(sequence).replaceAll("");\r
 168                 final Pattern othernonSeqChars = Pattern.compile("[_-]+");\r
 169                 sequence = othernonSeqChars.matcher(sequence).replaceAll("");\r
 170                 return sequence;\r
 171         }\r
 172 \r
 173         /**\r
 174          * @param sequence\r
 175          * @return true is the sequence is a protein sequence, false overwise\r
 176          */\r
 177         public static boolean isProteinSequence(String sequence) {\r
 178                 sequence = SequenceUtil.cleanSequence(sequence);\r
 179                 if (SequenceUtil.isNonAmbNucleotideSequence(sequence)) {\r
 180                         return false;\r
 181                 }\r
 182                 if (SequenceUtil.DIGIT.matcher(sequence).find()) {\r
 183                         return false;\r
 184                 }\r
 185                 if (SequenceUtil.NON_AA.matcher(sequence).find()) {\r
 186                         return false;\r
 187                 }\r
 188                 final Matcher protmatcher = SequenceUtil.AA.matcher(sequence);\r
 189                 return protmatcher.find();\r
 190         }\r
 191 \r
 192         /**\r
 193          * Check whether the sequence confirms to amboguous protein sequence\r
 194          * \r
 195          * @param sequence\r
 196          * @return return true only if the sequence if ambiguous protein sequence\r
 197          *         Return false otherwise. e.g. if the sequence is non-ambiguous\r
 198          *         protein or DNA\r
 199          */\r
 200         public static boolean isAmbiguosProtein(String sequence) {\r
 201                 sequence = SequenceUtil.cleanSequence(sequence);\r
 202                 if (SequenceUtil.isNonAmbNucleotideSequence(sequence)) {\r
 203                         return false;\r
 204                 }\r
 205                 if (SequenceUtil.DIGIT.matcher(sequence).find()) {\r
 206                         return false;\r
 207                 }\r
 208                 if (SequenceUtil.NON_AA.matcher(sequence).find()) {\r
 209                         return false;\r
 210                 }\r
 211                 if (SequenceUtil.AA.matcher(sequence).find()) {\r
 212                         return false;\r
 213                 }\r
 214                 final Matcher amb_prot = SequenceUtil.AMBIGUOUS_AA.matcher(sequence);\r
 215                 return amb_prot.find();\r
 216         }\r
 217 \r
 218         /**\r
 219          * Writes list of FastaSequeces into the outstream formatting the sequence\r
 220          * so that it contains width chars on each line\r
 221          * \r
 222          * @param outstream\r
 223          * @param sequences\r
 224          * @param width\r
 225          *            - the maximum number of characters to write in one line\r
 226          * @throws IOException\r
 227          */\r
 228         public static void writeFasta(final OutputStream outstream,\r
 229                         final List<FastaSequence> sequences, final int width)\r
 230                         throws IOException {\r
 231                 writeFastaKeepTheStream(outstream, sequences, width);\r
 232                 outstream.close();\r
 233         }\r
 234 \r
 235         public static void writeFastaKeepTheStream(final OutputStream outstream,\r
 236                         final List<FastaSequence> sequences, final int width)\r
 237                         throws IOException {\r
 238                 final OutputStreamWriter writer = new OutputStreamWriter(outstream);\r
 239                 final BufferedWriter fastawriter = new BufferedWriter(writer);\r
 240                 for (final FastaSequence fs : sequences) {\r
 241                         fastawriter.write(">" + fs.getId() + "\n");\r
 242                         fastawriter.write(fs.getFormatedSequence(width));\r
 243                         fastawriter.write("\n");\r
 244                 }\r
 245                 fastawriter.flush();\r
 246                 writer.flush();\r
 247         }\r
 248 \r
 249         /**\r
 250          * Reads fasta sequences from inStream into the list of FastaSequence\r
 251          * objects\r
 252          * \r
 253          * @param inStream\r
 254          *            from\r
 255          * @return list of FastaSequence objects\r
 256          * @throws IOException\r
 257          */\r
 258         public static List<FastaSequence> readFasta(final InputStream inStream)\r
 259                         throws IOException {\r
 260                 final List<FastaSequence> seqs = new ArrayList<FastaSequence>();\r
 261 \r
 262                 final BufferedReader infasta = new BufferedReader(\r
 263                                 new InputStreamReader(inStream, "UTF8"), 16000);\r
 264                 final Pattern pattern = Pattern.compile("//s+");\r
 265 \r
 266                 String line;\r
 267                 String sname = "", seqstr = null;\r
 268                 do {\r
 269                         line = infasta.readLine();\r
 270                         if ((line == null) || line.startsWith(">")) {\r
 271                                 if (seqstr != null) {\r
 272                                         seqs.add(new FastaSequence(sname.substring(1), seqstr));\r
 273                                 }\r
 274                                 sname = line; // remove >\r
 275                                 seqstr = "";\r
 276                         } else {\r
 277                                 final String subseq = pattern.matcher(line).replaceAll("");\r
 278                                 seqstr += subseq;\r
 279                         }\r
 280                 } while (line != null);\r
 281 \r
 282                 infasta.close();\r
 283                 return seqs;\r
 284         }\r
 285 \r
 286         /**\r
 287          * Writes FastaSequence in the file, each sequence will take one line only\r
 288          * \r
 289          * @param os\r
 290          * @param sequences\r
 291          * @throws IOException\r
 292          */\r
 293         public static void writeFasta(final OutputStream os,\r
 294                         final List<FastaSequence> sequences) throws IOException {\r
 295                 final OutputStreamWriter outWriter = new OutputStreamWriter(os);\r
 296                 final BufferedWriter fasta_out = new BufferedWriter(outWriter);\r
 297                 for (final FastaSequence fs : sequences) {\r
 298                         fasta_out.write(fs.getOnelineFasta());\r
 299                 }\r
 300                 fasta_out.close();\r
 301                 outWriter.close();\r
 302         }\r
 303 \r
 304         public static Map<String, Score> readJRonn(final File result)\r
 305                         throws IOException, UnknownFileFormatException {\r
 306                 InputStream input = new FileInputStream(result);\r
 307                 Map<String, Score> sequences = readJRonn(input);\r
 308                 input.close();\r
 309                 return sequences;\r
 310         }\r
 311 \r
 312         /**\r
 313          * Reader for JRonn horizontal file format\r
 314          * \r
 315          * <pre>\r
 316          * &gtFoobar M G D T T A G 0.48 0.42\r
 317          * 0.42 0.48 0.52 0.53 0.54\r
 318          * \r
 319          * <pre>\r
 320          * Where all values are tab delimited\r
 321          * \r
 322          * @param inStream\r
 323          *            the InputStream connected to the JRonn output file\r
 324          * @return List of {@link AnnotatedSequence} objects\r
 325          * @throws IOException\r
 326          *             is thrown if the inStream has problems accessing the data\r
 327          * @throws UnknownFileFormatException\r
 328          *             is thrown if the inStream represents an unknown source of\r
 329          * data, i.e. not a JRonn output\r
 330          */\r
 331         public static Map<String, Score> readJRonn(final InputStream inStream)\r
 332                         throws IOException, UnknownFileFormatException {\r
 333                 final Map<String, Score> seqs = new HashMap<String, Score>();\r
 334 \r
 335                 final BufferedReader infasta = new BufferedReader(\r
 336                                 new InputStreamReader(inStream, "UTF8"), 16000);\r
 337 \r
 338                 String line;\r
 339                 String sname = "";\r
 340                 do {\r
 341                         line = infasta.readLine();\r
 342                         if (line == null || line.isEmpty()) {\r
 343                                 // skip empty lines\r
 344                                 continue;\r
 345                         }\r
 346                         if (line.startsWith(">")) {\r
 347                                 // read name\r
 348                                 sname = line.trim().substring(1);\r
 349                                 // read sequence line\r
 350                                 line = infasta.readLine();\r
 351                                 final String sequence = line.replace("\t", "");\r
 352                                 // read annotation line\r
 353                                 line = infasta.readLine();\r
 354                                 String[] annotValues = line.split("\t");\r
 355                                 float[] annotation = convertToNumber(annotValues);\r
 356                                 if (annotation.length != sequence.length()) {\r
 357                                         throw new UnknownFileFormatException(\r
 358                                                         "File does not look like Jronn horizontally formatted output file!\n"\r
 359                                                                         + JRONN_WRONG_FORMAT_MESSAGE);\r
 360                                 }\r
 361                                 seqs.put(sname, new Score(DisorderMethod.JRonn, annotation));\r
 362                         }\r
 363                 } while (line != null);\r
 364 \r
 365                 infasta.close();\r
 366                 return seqs;\r
 367         }\r
 368         private static float[] convertToNumber(String[] annotValues)\r
 369                         throws UnknownFileFormatException {\r
 370                 float[] annotation = new float[annotValues.length];\r
 371                 try {\r
 372                         for (int i = 0; i < annotation.length; i++) {\r
 373                                 annotation[i] = Float.parseFloat(annotValues[i]);\r
 374                         }\r
 375                 } catch (NumberFormatException e) {\r
 376                         throw new UnknownFileFormatException(JRONN_WRONG_FORMAT_MESSAGE,\r
 377                                         e.getCause());\r
 378                 }\r
 379                 return annotation;\r
 380         }\r
 381 \r
 382         private static final String JRONN_WRONG_FORMAT_MESSAGE = "Jronn file must be in the following format:\n"\r
 383                         + ">sequence_name\n "\r
 384                         + "M    V       S\n"\r
 385                         + "0.43 0.22    0.65\n"\r
 386                         + "Where first line is the sequence name,\n"\r
 387                         + "second line is the tab delimited sequence,\n"\r
 388                         + "third line contains tab delimited disorder prediction values.\n"\r
 389                         + "No lines are allowed between these three. Additionally, the number of  "\r
 390                         + "sequence residues must be equal to the number of the disorder values.";\r
 391 \r
 392         /**\r
 393          * Closes the Closable and logs the exception if any\r
 394          * \r
 395          * @param log\r
 396          * @param stream\r
 397          */\r
 398         public final static void closeSilently(java.util.logging.Logger log,\r
 399                         Closeable stream) {\r
 400                 if (stream != null) {\r
 401                         try {\r
 402                                 stream.close();\r
 403                         } catch (IOException e) {\r
 404                                 log.log(Level.WARNING, e.getLocalizedMessage(), e.getCause());\r
 405                         }\r
 406                 }\r
 407         }\r
 408 \r
 409         /**\r
 410          * \r
 411          > Foobar_dundeefriends\r
 412          * \r
 413          * # COILS 34-41, 50-58, 83-91, 118-127, 160-169, 191-220, 243-252, 287-343\r
 414          * \r
 415          * # REM465 355-368\r
 416          * \r
 417          * # HOTLOOPS 190-204\r
 418          * \r
 419          * # RESIDUE COILS REM465 HOTLOOPS\r
 420          * \r
 421          * M 0.86010 0.88512 0.37094\r
 422          * \r
 423          * T 0.79983 0.85864 0.44331\r
 424          * \r
 425          * >Next Sequence name\r
 426          * \r
 427          * \r
 428          * @param input\r
 429          * @return\r
 430          * @throws IOException\r
 431          * @throws UnknownFileFormatException\r
 432          */\r
 433         public static Map<FastaSequence, Set<Score>> readDisembl(\r
 434                         final InputStream input) throws IOException,\r
 435                         UnknownFileFormatException {\r
 436                 Scanner scan = new Scanner(input);\r
 437                 scan.useDelimiter(">");\r
 438                 if (!scan.hasNext()) {\r
 439                         throw new UnknownFileFormatException(\r
 440                                         "In Disembl score format each sequence score is expected "\r
 441                                                         + "to start from the line: >Sequence name "\r
 442                                                         + " No such line was found!");\r
 443                 }\r
 444 \r
 445                 Map<FastaSequence, Set<Score>> results = new HashMap<FastaSequence, Set<Score>>();\r
 446                 int seqCounter = 0;\r
 447                 while (scan.hasNext()) {\r
 448                         seqCounter++;\r
 449                         String singleSeq = scan.next();\r
 450                         Scanner scansingle = new Scanner(singleSeq);\r
 451                         if (!scansingle.hasNextLine()) {\r
 452                                 throw new RuntimeException(\r
 453                                                 "The input looks like an incomplete disembl file - cannot parse!");\r
 454                         }\r
 455 \r
 456                         StringBuffer seqbuffer = new StringBuffer();\r
 457                         ArrayList<Float> coils = new ArrayList<Float>();\r
 458                         ArrayList<Float> rem = new ArrayList<Float>();\r
 459                         ArrayList<Float> hotloops = new ArrayList<Float>();\r
 460 \r
 461                         String sequenceName = scansingle.nextLine().trim();\r
 462                         SortedSet<Range> coilsR = parseRanges(DisemblResult.COILS,\r
 463                                         scansingle.nextLine());\r
 464                         SortedSet<Range> rem465R = parseRanges(DisemblResult.REM465,\r
 465                                         scansingle.nextLine());\r
 466                         SortedSet<Range> loopsR = parseRanges(DisemblResult.HOTLOOPS,\r
 467                                         scansingle.nextLine());\r
 468 \r
 469                         String title = scansingle.nextLine();\r
 470                         assert title.startsWith("# RESIDUE COILS REM465 HOTLOOPS") : ">Sequence_name must follow column title: # RESIDUE COILS REM465 HOTLOOPS!";\r
 471 \r
 472                         while (scansingle.hasNext()) {\r
 473                                 seqbuffer.append(scansingle.next());\r
 474                                 coils.add(scansingle.nextFloat());\r
 475                                 rem.add(scansingle.nextFloat());\r
 476                                 hotloops.add(scansingle.nextFloat());\r
 477                         }\r
 478                         FastaSequence fs = new FastaSequence(sequenceName,\r
 479                                         seqbuffer.toString());\r
 480                         Set<Score> scores = new HashSet<Score>();\r
 481                         scores.add(new Score(DisemblResult.COILS, coils, coilsR));\r
 482                         scores.add(new Score(DisemblResult.HOTLOOPS, hotloops, rem465R));\r
 483                         scores.add(new Score(DisemblResult.REM465, rem, loopsR));\r
 484                         results.put(fs, scores);\r
 485 \r
 486                         scansingle.close();\r
 487                 }\r
 488 \r
 489                 input.close();\r
 490                 return results;\r
 491         }\r
 492 \r
 493         /**\r
 494          * Parsing:\r
 495          * \r
 496          * # COILS 34-41, 50-58, 83-91, 118-127, 160-169, 191-220, 243-252, 287-343,\r
 497          * 350-391, 429-485, 497-506, 539-547\r
 498          * \r
 499          * # REM465 355-368\r
 500          * \r
 501          * # HOTLOOPS 190-204\r
 502          * \r
 503          * @param lines\r
 504          * @return\r
 505          */\r
 506         private static SortedSet<Range> parseRanges(Enum resultType, String lines) {\r
 507                 SortedSet<Range> ranges = new TreeSet<Range>();\r
 508                 Scanner scan = new Scanner(lines);\r
 509 \r
 510                 assert scan.hasNext();\r
 511                 assert "#".equals(scan.next()); // pass delimiter #\r
 512                 String type = scan.next(); // pass enum name e.g. COILS\r
 513                 assert resultType.toString().equalsIgnoreCase(type) : "Unknown result type: "\r
 514                                 + resultType.toString();\r
 515 \r
 516                 // beginning of the ranges\r
 517                 scan.useDelimiter(",");\r
 518                 while (scan.hasNext()) {\r
 519                         String range = scan.next();\r
 520                         if (!Util.isEmpty(range)) {\r
 521                                 ranges.add(new Range(range.split("-")));\r
 522                         }\r
 523                 }\r
 524                 return ranges;\r
 525         }\r
 526 \r
 527         public static Map<String, Set<Score>> removeSequences(\r
 528                         Map<FastaSequence, Set<Score>> disemblResults) {\r
 529                 Map<String, Set<Score>> seqNameScores = new HashMap<String, Set<Score>>();\r
 530                 for (Map.Entry<FastaSequence, Set<Score>> dres : disemblResults\r
 531                                 .entrySet()) {\r
 532                         seqNameScores.put(dres.getKey().getId(), dres.getValue());\r
 533                 }\r
 534                 return seqNameScores;\r
 535         }\r
 536 \r
 537         /**\r
 538          * \r
 539          > Foobar_dundeefriends\r
 540          * \r
 541          * # COILS 34-41, 50-58, 83-91, 118-127, 160-169, 191-220, 243-252, 287-343\r
 542          * \r
 543          * # REM465 355-368\r
 544          * \r
 545          * # HOTLOOPS 190-204\r
 546          * \r
 547          * # RESIDUE COILS REM465 HOTLOOPS\r
 548          * \r
 549          * M 0.86010 0.88512 0.37094\r
 550          * \r
 551          * T 0.79983 0.85864 0.44331\r
 552          * \r
 553          * >Next Sequence name\r
 554          * \r
 555          * \r
 556          * @param input\r
 557          * @return\r
 558          * @throws IOException\r
 559          * @throws UnknownFileFormatException\r
 560          */\r
 561         public static Map<FastaSequence, Set<Score>> readGlobPlot(\r
 562                         final InputStream input) throws IOException,\r
 563                         UnknownFileFormatException {\r
 564                 Scanner scan = new Scanner(input);\r
 565                 scan.useDelimiter(">");\r
 566                 if (!scan.hasNext()) {\r
 567                         throw new UnknownFileFormatException(\r
 568                                         "In GlobPlot score format each sequence score is expected "\r
 569                                                         + "to start from the line: >Sequence name "\r
 570                                                         + " No such line was found!");\r
 571                 }\r
 572 \r
 573                 Map<FastaSequence, Set<Score>> results = new HashMap<FastaSequence, Set<Score>>();\r
 574                 int seqCounter = 0;\r
 575                 while (scan.hasNext()) {\r
 576                         seqCounter++;\r
 577                         String singleSeq = scan.next();\r
 578                         Scanner scansingle = new Scanner(singleSeq);\r
 579                         if (!scansingle.hasNextLine()) {\r
 580                                 throw new RuntimeException(\r
 581                                                 "The input looks like an incomplete GlobPlot file - cannot parse!");\r
 582                         }\r
 583 \r
 584                         StringBuffer seqbuffer = new StringBuffer();\r
 585                         ArrayList<Float> dydxScore = new ArrayList<Float>();\r
 586                         ArrayList<Float> rawScore = new ArrayList<Float>();\r
 587                         ArrayList<Float> smoothedScore = new ArrayList<Float>();\r
 588 \r
 589                         String sequenceName = scansingle.nextLine().trim();\r
 590                         SortedSet<Range> domsR = parseRanges(GlobProtResult.GlobDoms,\r
 591                                         scansingle.nextLine());\r
 592                         SortedSet<Range> disorderR = parseRanges(GlobProtResult.Disorder,\r
 593                                         scansingle.nextLine());\r
 594 \r
 595                         String title = scansingle.nextLine();\r
 596                         assert title.startsWith("# RESIDUE      DYDX") : ">Sequence_name must follow column title: # RESIDUE DYDX RAW SMOOTHED!";\r
 597 \r
 598                         while (scansingle.hasNext()) {\r
 599                                 seqbuffer.append(scansingle.next());\r
 600                                 dydxScore.add(scansingle.nextFloat());\r
 601                                 rawScore.add(scansingle.nextFloat());\r
 602                                 smoothedScore.add(scansingle.nextFloat());\r
 603                         }\r
 604                         FastaSequence fs = new FastaSequence(sequenceName,\r
 605                                         seqbuffer.toString());\r
 606                         Set<Score> scores = new HashSet<Score>();\r
 607                         scores.add(new Score(GlobProtResult.Disorder, disorderR));\r
 608                         scores.add(new Score(GlobProtResult.GlobDoms, domsR));\r
 609                         scores.add(new Score(GlobProtResult.Dydx, dydxScore));\r
 610                         scores.add(new Score(GlobProtResult.RawScore, rawScore));\r
 611                         scores.add(new Score(GlobProtResult.SmoothedScore, smoothedScore));\r
 612                         results.put(fs, scores);\r
 613 \r
 614                         scansingle.close();\r
 615                 }\r
 616 \r
 617                 input.close();\r
 618                 return results;\r
 619         }\r
 620         /**\r
 621          * Read AACon result with no alignment files. This method leaves incoming\r
 622          * the InputStream results open!\r
 623          * \r
 624          * @param results\r
 625          *            output file of AAConservation\r
 626          * @return Map with keys {@link ConservationMethod} -> float[]\r
 627          */\r
 628         public static HashSet<Score> readAAConResults(InputStream results) {\r
 629                 if (results == null) {\r
 630                         throw new NullPointerException(\r
 631                                         "InputStream with results must be provided");\r
 632                 }\r
 633                 HashSet<Score> annotations = new HashSet<Score>();\r
 634                 Scanner sc = new Scanner(results);\r
 635                 sc.useDelimiter("#");\r
 636                 while (sc.hasNext()) {\r
 637                         String line = sc.next();\r
 638                         int spacePos = line.indexOf(" ");\r
 639                         assert spacePos > 0 : "Space is expected as delimited between method "\r
 640                                         + "name and values!";\r
 641                         String methodLine = line.substring(0, spacePos);\r
 642                         ConservationMethod method = ConservationMethod\r
 643                                         .getMethod(methodLine);\r
 644                         assert method != null : "Method " + methodLine\r
 645                                         + " is not recognized! ";\r
 646                         Scanner valuesScanner = new Scanner(line.substring(spacePos));\r
 647                         ArrayList<Float> values = new ArrayList<Float>();\r
 648                         while (valuesScanner.hasNextDouble()) {\r
 649                                 Double value = valuesScanner.nextDouble();\r
 650                                 values.add(value.floatValue());\r
 651                         }\r
 652                         annotations.add(new Score(method, values));\r
 653                 }\r
 654                 return annotations;\r
 655         }\r
 656 \r
 657         /**\r
 658          * Reads and parses Fasta or Clustal formatted file into a list of\r
 659          * FastaSequence objects\r
 660          * \r
 661          * @param inFilePath\r
 662          *            the path to the input file\r
 663          * @throws IOException\r
 664          *             if the file denoted by inFilePath cannot be read\r
 665          * @throws UnknownFileFormatException\r
 666          *             if the inFilePath points to the file which format cannot be\r
 667          *             recognised\r
 668          * @return the List of FastaSequence objects\r
 669          * \r
 670          */\r
 671         public static List<FastaSequence> openInputStream(String inFilePath)\r
 672                         throws IOException, UnknownFileFormatException {\r
 673 \r
 674                 // This stream gets closed in isValidClustalFile method\r
 675                 InputStream inStrForValidation = new FileInputStream(inFilePath);\r
 676                 // This stream is closed in the calling methods\r
 677                 InputStream inStr = new FileInputStream(inFilePath);\r
 678                 List<FastaSequence> fastaSeqs = null;\r
 679                 if (ClustalAlignmentUtil.isValidClustalFile(inStrForValidation)) {\r
 680                         Alignment al = ClustalAlignmentUtil.readClustalFile(inStr);\r
 681                         // alignment cannot be null see\r
 682                         // ClustalAlignmentUtil.readClustalFile(inStr);\r
 683                         fastaSeqs = al.getSequences();\r
 684                 } else {\r
 685                         fastaSeqs = SequenceUtil.readFasta(inStr);\r
 686                 }\r
 687                 return fastaSeqs;\r
 688         }\r
 689 \r
 690 }\r
 691 \r
 692 enum DisemblResult {\r
 693         /** These contains ranges and scores */\r
 694         COILS, REM465, HOTLOOPS\r
 695 }\r
 696 enum GlobProtResult {\r
 697         /** This a range with no scores */\r
 698         GlobDoms,\r
 699         /** This a range with no scores */\r
 700         Disorder,\r
 701         /** This a score with no range */\r
 702         Dydx,\r
 703         /** This a score with no range */\r
 704         SmoothedScore,\r
 705         /** This a score with no range */\r
 706         RawScore\r
 707 }