/* * Jalview - A Sequence Alignment Editor and Viewer * Copyright (C) 2007 AM Waterhouse, J Procter, G Barton, M Clamp, S Searle * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public License * as published by the Free Software Foundation; either version 2 * of the License, or (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA */ package jalview.analysis; import java.util.*; import jalview.datamodel.*; /** *

Title:

* *

Description:

* *

Copyright: Copyright (c) 2004

* *

Company: Dundee University

* * @author not attributable * @version 1.0 */ public class SeqsetUtils { /** * Store essential properties of a sequence in a hashtable for later recovery * Keys are Name, Start, End, SeqFeatures, PdbId * @param seq SequenceI * @return Hashtable */ public static Hashtable SeqCharacterHash(SequenceI seq) { Hashtable sqinfo = new Hashtable(); sqinfo.put("Name", seq.getName()); sqinfo.put("Start", new Integer(seq.getStart())); sqinfo.put("End", new Integer(seq.getEnd())); if (seq.getDescription() != null) { sqinfo.put("Description", seq.getDescription()); } Vector sfeat = new Vector(); jalview.datamodel.SequenceFeature[] sfarray = seq.getSequenceFeatures(); if (sfarray != null && sfarray.length > 0) { for (int i = 0; i < sfarray.length; i++) { sfeat.add(sfarray[i]); } } sqinfo.put("SeqFeatures", sfeat); sqinfo.put("PdbId", (seq.getPDBId() != null) ? seq.getPDBId() : new Vector()); sqinfo.put("datasetSequence", (seq.getDatasetSequence() != null) ? seq.getDatasetSequence() : new Sequence("THISISAPLACEHOLDER", "")); return sqinfo; } /** * Recover essential properties of a sequence from a hashtable * TODO: replace these methods with something more elegant. * @param sq SequenceI * @param sqinfo Hashtable * @return boolean true if name was not updated from sqinfo Name entry */ public static boolean SeqCharacterUnhash(SequenceI sq, Hashtable sqinfo) { boolean namePresent = true; if (sqinfo == null) { return false; } String oldname = (String) sqinfo.get("Name"); Integer start = (Integer) sqinfo.get("Start"); Integer end = (Integer) sqinfo.get("End"); Vector sfeatures = (Vector) sqinfo.get( "SeqFeatures"); Vector pdbid = (Vector) sqinfo.get("PdbId"); String description = (String) sqinfo.get("Description"); Sequence seqds = (Sequence) sqinfo.get("datasetSequence"); if (oldname == null) { namePresent = false; } else { sq.setName(oldname); } if (pdbid != null && pdbid.size() > 0) { sq.setPDBId(pdbid); } if ( (start != null) && (end != null)) { sq.setStart(start.intValue()); sq.setEnd(end.intValue()); } if ( (sfeatures != null) && (sfeatures.size() > 0)) { SequenceFeature[] sfarray = (SequenceFeature[]) sfeatures.toArray(); sq.setSequenceFeatures(sfarray); } if (description != null) { sq.setDescription(description); } if ( (seqds != null) && ! (seqds.getName().equals("THISISAPLACEHOLDER") && seqds.getLength() == 0)) { sq.setDatasetSequence(seqds); } return namePresent; } /** * Form of the unique name used in uniquify for the i'th sequence in an ordered vector of sequences. * @param i int * @return String */ public static String unique_name(int i) { return new String("Sequence" + i); } /** * Generates a hash of SeqCharacterHash properties for each sequence * in a sequence set, and optionally renames the sequences to an * unambiguous 'safe' name. * @param sequences SequenceI[] * @param write_names boolean set this to rename each of the sequences to its unique_name(index) name * @return Hashtable to be passed to @see deuniquify to recover original names (and properties) for renamed sequences */ public static Hashtable uniquify(SequenceI[] sequences, boolean write_names) { // Generate a safely named sequence set and a hash to recover the sequence names Hashtable map = new Hashtable(); //String[] un_names = new String[sequences.length]; for (int i = 0; i < sequences.length; i++) { String safename = unique_name(i); map.put(safename, SeqCharacterHash(sequences[i])); if (write_names) { sequences[i].setName(safename); } } return map; } /** * recover unsafe sequence names and original properties for a sequence * set using a map generated by @see uniquify(sequences,true) * @param map Hashtable * @param sequences SequenceI[] * @return boolean */ public static boolean deuniquify(Hashtable map, SequenceI[] sequences) { jalview.analysis.SequenceIdMatcher matcher = new SequenceIdMatcher( sequences); SequenceI msq = null; Enumeration keys = map.keys(); Vector unmatched = new Vector(); for (int i = 0, j = sequences.length; i < j; i++) { unmatched.add(sequences[i]); } while (keys.hasMoreElements()) { Object key = keys.nextElement(); if (key instanceof String) { if ( (msq = matcher.findIdMatch( (String) key)) != null) { Hashtable sqinfo = (Hashtable) map.get(key); unmatched.remove(msq); SeqCharacterUnhash(msq, sqinfo); } else { System.err.println("Can't find '" + ( (String) key) + "' in uniquified alignment"); } } } if (unmatched.size() > 0) { System.err.println("Did not find matches for :"); for (Enumeration i = unmatched.elements(); i.hasMoreElements(); System.out.println( ( (SequenceI) i.nextElement()).getName())) { ; } return false; } return true; } /** * returns a subset of the sequenceI seuqences, * including only those that contain at least one residue. * @param sequences SequenceI[] * @return SequenceI[] */ public static SequenceI[] getNonEmptySequenceSet(SequenceI[] sequences) { // Identify first row of alignment with residues for prediction boolean ungapped[] = new boolean[sequences.length]; int msflen = 0; for (int i = 0, j = sequences.length; i < j; i++) { String tempseq = jalview.analysis.AlignSeq.extractGaps( jalview.util.Comparison.GapChars, sequences[i].getSequenceAsString()); if (tempseq.length() == 0) { ungapped[i] = false; } else { ungapped[i] = true; msflen++; } } if (msflen == 0) { return null; // no minimal set } // compose minimal set SequenceI[] mset = new SequenceI[msflen]; for (int i = 0, j = sequences.length, k = 0; i < j; i++) { if (ungapped[i]) { mset[k++] = sequences[i]; } } ungapped = null; return mset; } }