[jalview.git] / src / jalview / analysis / AlignmentUtils.java

/*
 * Jalview - A Sequence Alignment Editor and Viewer ($$Version-Rel$$)
 * Copyright (C) $$Year-Rel$$ The Jalview Authors
 * 
 * This file is part of Jalview.
 * 
 * Jalview 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 3
 * of the License, or (at your option) any later version.
 *  
 * Jalview 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 Jalview.  If not, see <http://www.gnu.org/licenses/>.
 * The Jalview Authors are detailed in the 'AUTHORS' file.
 */
package jalview.analysis;

import jalview.datamodel.AlignedCodon;
import jalview.datamodel.AlignedCodonFrame;
import jalview.datamodel.Alignment;
import jalview.datamodel.AlignmentAnnotation;
import jalview.datamodel.AlignmentI;
import jalview.datamodel.DBRefEntry;
import jalview.datamodel.IncompleteCodonException;
import jalview.datamodel.Mapping;
import jalview.datamodel.Sequence;
import jalview.datamodel.SequenceFeature;
import jalview.datamodel.SequenceGroup;
import jalview.datamodel.SequenceI;
import jalview.io.gff.SequenceOntologyFactory;
import jalview.io.gff.SequenceOntologyI;
import jalview.schemes.ResidueProperties;
import jalview.util.Comparison;
import jalview.util.MapList;
import jalview.util.MappingUtils;
import jalview.util.StringUtils;

import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.Comparator;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.NoSuchElementException;
import java.util.Set;
import java.util.TreeMap;

/**
 * grab bag of useful alignment manipulation operations Expect these to be
 * refactored elsewhere at some point.
 * 
 * @author jimp
 * 
 */
public class AlignmentUtils
{

  /**
   * given an existing alignment, create a new alignment including all, or up to
   * flankSize additional symbols from each sequence's dataset sequence
   * 
   * @param core
   * @param flankSize
   * @return AlignmentI
   */
  public static AlignmentI expandContext(AlignmentI core, int flankSize)
  {
    List<SequenceI> sq = new ArrayList<SequenceI>();
    int maxoffset = 0;
    for (SequenceI s : core.getSequences())
    {
      SequenceI newSeq = s.deriveSequence();
      final int newSeqStart = newSeq.getStart() - 1;
      if (newSeqStart > maxoffset
              && newSeq.getDatasetSequence().getStart() < s.getStart())
      {
        maxoffset = newSeqStart;
      }
      sq.add(newSeq);
    }
    if (flankSize > -1)
    {
      maxoffset = Math.min(maxoffset, flankSize);
    }

    /*
     * now add offset left and right to create an expanded alignment
     */
    for (SequenceI s : sq)
    {
      SequenceI ds = s;
      while (ds.getDatasetSequence() != null)
      {
        ds = ds.getDatasetSequence();
      }
      int s_end = s.findPosition(s.getStart() + s.getLength());
      // find available flanking residues for sequence
      int ustream_ds = s.getStart() - ds.getStart();
      int dstream_ds = ds.getEnd() - s_end;

      // build new flanked sequence

      // compute gap padding to start of flanking sequence
      int offset = maxoffset - ustream_ds;

      // padding is gapChar x ( maxoffset - min(ustream_ds, flank)
      if (flankSize >= 0)
      {
        if (flankSize < ustream_ds)
        {
          // take up to flankSize residues
          offset = maxoffset - flankSize;
          ustream_ds = flankSize;
        }
        if (flankSize <= dstream_ds)
        {
          dstream_ds = flankSize - 1;
        }
      }
      // TODO use Character.toLowerCase to avoid creating String objects?
      char[] upstream = new String(ds.getSequence(s.getStart() - 1
              - ustream_ds, s.getStart() - 1)).toLowerCase().toCharArray();
      char[] downstream = new String(ds.getSequence(s_end - 1, s_end
              + dstream_ds)).toLowerCase().toCharArray();
      char[] coreseq = s.getSequence();
      char[] nseq = new char[offset + upstream.length + downstream.length
              + coreseq.length];
      char c = core.getGapCharacter();

      int p = 0;
      for (; p < offset; p++)
      {
        nseq[p] = c;
      }

      System.arraycopy(upstream, 0, nseq, p, upstream.length);
      System.arraycopy(coreseq, 0, nseq, p + upstream.length,
              coreseq.length);
      System.arraycopy(downstream, 0, nseq, p + coreseq.length
              + upstream.length, downstream.length);
      s.setSequence(new String(nseq));
      s.setStart(s.getStart() - ustream_ds);
      s.setEnd(s_end + downstream.length);
    }
    AlignmentI newAl = new jalview.datamodel.Alignment(
            sq.toArray(new SequenceI[0]));
    for (SequenceI s : sq)
    {
      if (s.getAnnotation() != null)
      {
        for (AlignmentAnnotation aa : s.getAnnotation())
        {
          aa.adjustForAlignment(); // JAL-1712 fix
          newAl.addAnnotation(aa);
        }
      }
    }
    newAl.setDataset(core.getDataset());
    return newAl;
  }

  /**
   * Returns the index (zero-based position) of a sequence in an alignment, or
   * -1 if not found.
   * 
   * @param al
   * @param seq
   * @return
   */
  public static int getSequenceIndex(AlignmentI al, SequenceI seq)
  {
    int result = -1;
    int pos = 0;
    for (SequenceI alSeq : al.getSequences())
    {
      if (alSeq == seq)
      {
        result = pos;
        break;
      }
      pos++;
    }
    return result;
  }

  /**
   * Returns a map of lists of sequences in the alignment, keyed by sequence
   * name. For use in mapping between different alignment views of the same
   * sequences.
   * 
   * @see jalview.datamodel.AlignmentI#getSequencesByName()
   */
  public static Map<String, List<SequenceI>> getSequencesByName(
          AlignmentI al)
  {
    Map<String, List<SequenceI>> theMap = new LinkedHashMap<String, List<SequenceI>>();
    for (SequenceI seq : al.getSequences())
    {
      String name = seq.getName();
      if (name != null)
      {
        List<SequenceI> seqs = theMap.get(name);
        if (seqs == null)
        {
          seqs = new ArrayList<SequenceI>();
          theMap.put(name, seqs);
        }
        seqs.add(seq);
      }
    }
    return theMap;
  }

  /**
   * Build mapping of protein to cDNA alignment. Mappings are made between
   * sequences where the cDNA translates to the protein sequence. Any new
   * mappings are added to the protein alignment. Returns true if any mappings
   * either already exist or were added, else false.
   * 
   * @param proteinAlignment
   * @param cdnaAlignment
   * @return
   */
  public static boolean mapProteinAlignmentToCdna(
          final AlignmentI proteinAlignment, final AlignmentI cdnaAlignment)
  {
    if (proteinAlignment == null || cdnaAlignment == null)
    {
      return false;
    }

    Set<SequenceI> mappedDna = new HashSet<SequenceI>();
    Set<SequenceI> mappedProtein = new HashSet<SequenceI>();

    /*
     * First pass - map sequences where cross-references exist. This include
     * 1-to-many mappings to support, for example, variant cDNA.
     */
    boolean mappingPerformed = mapProteinToCdna(proteinAlignment,
            cdnaAlignment, mappedDna, mappedProtein, true);

    /*
     * Second pass - map sequences where no cross-references exist. This only
     * does 1-to-1 mappings and assumes corresponding sequences are in the same
     * order in the alignments.
     */
    mappingPerformed |= mapProteinToCdna(proteinAlignment, cdnaAlignment,
            mappedDna, mappedProtein, false);
    return mappingPerformed;
  }

  /**
   * Make mappings between compatible sequences (where the cDNA translation
   * matches the protein).
   * 
   * @param proteinAlignment
   * @param cdnaAlignment
   * @param mappedDna
   *          a set of mapped DNA sequences (to add to)
   * @param mappedProtein
   *          a set of mapped Protein sequences (to add to)
   * @param xrefsOnly
   *          if true, only map sequences where xrefs exist
   * @return
   */
  protected static boolean mapProteinToCdna(
          final AlignmentI proteinAlignment,
          final AlignmentI cdnaAlignment, Set<SequenceI> mappedDna,
          Set<SequenceI> mappedProtein, boolean xrefsOnly)
  {
    boolean mappingExistsOrAdded = false;
    List<SequenceI> thisSeqs = proteinAlignment.getSequences();
    for (SequenceI aaSeq : thisSeqs)
    {
      boolean proteinMapped = false;
      AlignedCodonFrame acf = new AlignedCodonFrame();

      for (SequenceI cdnaSeq : cdnaAlignment.getSequences())
      {
        /*
         * Always try to map if sequences have xref to each other; this supports
         * variant cDNA or alternative splicing for a protein sequence.
         * 
         * If no xrefs, try to map progressively, assuming that alignments have
         * mappable sequences in corresponding order. These are not
         * many-to-many, as that would risk mixing species with similar cDNA
         * sequences.
         */
        if (xrefsOnly && !AlignmentUtils.haveCrossRef(aaSeq, cdnaSeq))
        {
          continue;
        }

        /*
         * Don't map non-xrefd sequences more than once each. This heuristic
         * allows us to pair up similar sequences in ordered alignments.
         */
        if (!xrefsOnly
                && (mappedProtein.contains(aaSeq) || mappedDna
                        .contains(cdnaSeq)))
        {
          continue;
        }
        if (mappingExists(proteinAlignment.getCodonFrames(),
                aaSeq.getDatasetSequence(), cdnaSeq.getDatasetSequence()))
        {
          mappingExistsOrAdded = true;
        }
        else
        {
          MapList map = mapCdnaToProtein(aaSeq, cdnaSeq);
          if (map != null)
          {
            acf.addMap(cdnaSeq, aaSeq, map);
            mappingExistsOrAdded = true;
            proteinMapped = true;
            mappedDna.add(cdnaSeq);
            mappedProtein.add(aaSeq);
          }
        }
      }
      if (proteinMapped)
      {
        proteinAlignment.addCodonFrame(acf);
      }
    }
    return mappingExistsOrAdded;
  }

  /**
   * Answers true if the mappings include one between the given (dataset)
   * sequences.
   */
  public static boolean mappingExists(List<AlignedCodonFrame> mappings,
          SequenceI aaSeq, SequenceI cdnaSeq)
  {
    if (mappings != null)
    {
      for (AlignedCodonFrame acf : mappings)
      {
        if (cdnaSeq == acf.getDnaForAaSeq(aaSeq))
        {
          return true;
        }
      }
    }
    return false;
  }

  /**
   * Builds a mapping (if possible) of a cDNA to a protein sequence.
   * <ul>
   * <li>first checks if the cdna translates exactly to the protein sequence</li>
   * <li>else checks for translation after removing a STOP codon</li>
   * <li>else checks for translation after removing a START codon</li>
   * <li>if that fails, inspect CDS features on the cDNA sequence</li>
   * </ul>
   * Returns null if no mapping is determined.
   * 
   * @param proteinSeq
   *          the aligned protein sequence
   * @param cdnaSeq
   *          the aligned cdna sequence
   * @return
   */
  public static MapList mapCdnaToProtein(SequenceI proteinSeq,
          SequenceI cdnaSeq)
  {
    /*
     * Here we handle either dataset sequence set (desktop) or absent (applet).
     * Use only the char[] form of the sequence to avoid creating possibly large
     * String objects.
     */
    final SequenceI proteinDataset = proteinSeq.getDatasetSequence();
    char[] aaSeqChars = proteinDataset != null ? proteinDataset
            .getSequence() : proteinSeq.getSequence();
    final SequenceI cdnaDataset = cdnaSeq.getDatasetSequence();
    char[] cdnaSeqChars = cdnaDataset != null ? cdnaDataset.getSequence()
            : cdnaSeq.getSequence();
    if (aaSeqChars == null || cdnaSeqChars == null)
    {
      return null;
    }

    /*
     * cdnaStart/End, proteinStartEnd are base 1 (for dataset sequence mapping)
     */
    final int mappedLength = 3 * aaSeqChars.length;
    int cdnaLength = cdnaSeqChars.length;
    int cdnaStart = cdnaSeq.getStart();
    int cdnaEnd = cdnaSeq.getEnd();
    final int proteinStart = proteinSeq.getStart();
    final int proteinEnd = proteinSeq.getEnd();

    /*
     * If lengths don't match, try ignoring stop codon (if present)
     */
    if (cdnaLength != mappedLength && cdnaLength > 2)
    {
      String lastCodon = String.valueOf(cdnaSeqChars, cdnaLength - 3, 3)
              .toUpperCase();
      for (String stop : ResidueProperties.STOP)
      {
        if (lastCodon.equals(stop))
        {
          cdnaEnd -= 3;
          cdnaLength -= 3;
          break;
        }
      }
    }

    /*
     * If lengths still don't match, try ignoring start codon.
     */
    int startOffset = 0;
    if (cdnaLength != mappedLength
            && cdnaLength > 2
            && String.valueOf(cdnaSeqChars, 0, 3).toUpperCase()
                    .equals(ResidueProperties.START))
    {
      startOffset += 3;
      cdnaStart += 3;
      cdnaLength -= 3;
    }

    if (translatesAs(cdnaSeqChars, startOffset, aaSeqChars))
    {
      /*
       * protein is translation of dna (+/- start/stop codons)
       */
      MapList map = new MapList(new int[] { cdnaStart, cdnaEnd }, new int[]
      { proteinStart, proteinEnd }, 3, 1);
      return map;
    }

    /*
     * translation failed - try mapping CDS annotated regions of dna
     */
    return mapCdsToProtein(cdnaSeq, proteinSeq);
  }

  /**
   * Test whether the given cdna sequence, starting at the given offset,
   * translates to the given amino acid sequence, using the standard translation
   * table. Designed to fail fast i.e. as soon as a mismatch position is found.
   * 
   * @param cdnaSeqChars
   * @param cdnaStart
   * @param aaSeqChars
   * @return
   */
  protected static boolean translatesAs(char[] cdnaSeqChars, int cdnaStart,
          char[] aaSeqChars)
  {
    if (cdnaSeqChars == null || aaSeqChars == null)
    {
      return false;
    }

    int aaPos = 0;
    int dnaPos = cdnaStart;
    for (; dnaPos < cdnaSeqChars.length - 2
            && aaPos < aaSeqChars.length; dnaPos += 3, aaPos++)
    {
      String codon = String.valueOf(cdnaSeqChars, dnaPos, 3);
      final String translated = ResidueProperties.codonTranslate(codon);

      /*
       * allow * in protein to match untranslatable in dna
       */
      final char aaRes = aaSeqChars[aaPos];
      if ((translated == null || "STOP".equals(translated)) && aaRes == '*')
      {
        continue;
      }
      if (translated == null || !(aaRes == translated.charAt(0)))
      {
        // debug
        // System.out.println(("Mismatch at " + i + "/" + aaResidue + ": "
        // + codon + "(" + translated + ") != " + aaRes));
        return false;
      }
    }

    /*
     * check we matched all of the protein sequence
     */
    if (aaPos != aaSeqChars.length)
    {
      return false;
    }

    /*
     * check we matched all of the dna except
     * for optional trailing STOP codon
     */
    if (dnaPos == cdnaSeqChars.length)
    {
      return true;
    }
    if (dnaPos == cdnaSeqChars.length - 3)
    {
      String codon = String.valueOf(cdnaSeqChars, dnaPos, 3);
      if ("STOP".equals(ResidueProperties.codonTranslate(codon)))
      {
        return true;
      }
    }
    return false;
  }

  /**
   * Align sequence 'seq' to match the alignment of a mapped sequence. Note this
   * currently assumes that we are aligning cDNA to match protein.
   * 
   * @param seq
   *          the sequence to be realigned
   * @param al
   *          the alignment whose sequence alignment is to be 'copied'
   * @param gap
   *          character string represent a gap in the realigned sequence
   * @param preserveUnmappedGaps
   * @param preserveMappedGaps
   * @return true if the sequence was realigned, false if it could not be
   */
  public static boolean alignSequenceAs(SequenceI seq, AlignmentI al,
          String gap, boolean preserveMappedGaps,
          boolean preserveUnmappedGaps)
  {
    /*
     * Get any mappings from the source alignment to the target (dataset)
     * sequence.
     */
    // TODO there may be one AlignedCodonFrame per dataset sequence, or one with
    // all mappings. Would it help to constrain this?
    List<AlignedCodonFrame> mappings = al.getCodonFrame(seq);
    if (mappings == null || mappings.isEmpty())
    {
      return false;
    }

    /*
     * Locate the aligned source sequence whose dataset sequence is mapped. We
     * just take the first match here (as we can't align like more than one
     * sequence).
     */
    SequenceI alignFrom = null;
    AlignedCodonFrame mapping = null;
    for (AlignedCodonFrame mp : mappings)
    {
      alignFrom = mp.findAlignedSequence(seq, al);
      if (alignFrom != null)
      {
        mapping = mp;
        break;
      }
    }

    if (alignFrom == null)
    {
      return false;
    }
    alignSequenceAs(seq, alignFrom, mapping, gap, al.getGapCharacter(),
            preserveMappedGaps, preserveUnmappedGaps);
    return true;
  }

  /**
   * Align sequence 'alignTo' the same way as 'alignFrom', using the mapping to
   * match residues and codons. Flags control whether existing gaps in unmapped
   * (intron) and mapped (exon) regions are preserved or not. Gaps between
   * intron and exon are only retained if both flags are set.
   * 
   * @param alignTo
   * @param alignFrom
   * @param mapping
   * @param myGap
   * @param sourceGap
   * @param preserveUnmappedGaps
   * @param preserveMappedGaps
   */
  public static void alignSequenceAs(SequenceI alignTo,
          SequenceI alignFrom, AlignedCodonFrame mapping, String myGap,
          char sourceGap, boolean preserveMappedGaps,
          boolean preserveUnmappedGaps)
  {
    // TODO generalise to work for Protein-Protein, dna-dna, dna-protein

    // aligned and dataset sequence positions, all base zero
    int thisSeqPos = 0;
    int sourceDsPos = 0;

    int basesWritten = 0;
    char myGapChar = myGap.charAt(0);
    int ratio = myGap.length();

    int fromOffset = alignFrom.getStart() - 1;
    int toOffset = alignTo.getStart() - 1;
    int sourceGapMappedLength = 0;
    boolean inExon = false;
    final char[] thisSeq = alignTo.getSequence();
    final char[] thatAligned = alignFrom.getSequence();
    StringBuilder thisAligned = new StringBuilder(2 * thisSeq.length);

    /*
     * Traverse the 'model' aligned sequence
     */
    for (char sourceChar : thatAligned)
    {
      if (sourceChar == sourceGap)
      {
        sourceGapMappedLength += ratio;
        continue;
      }

      /*
       * Found a non-gap character. Locate its mapped region if any.
       */
      sourceDsPos++;
      // Note mapping positions are base 1, our sequence positions base 0
      int[] mappedPos = mapping.getMappedRegion(alignTo, alignFrom,
              sourceDsPos + fromOffset);
      if (mappedPos == null)
      {
        /*
         * unmapped position; treat like a gap
         */
        sourceGapMappedLength += ratio;
        // System.err.println("Can't align: no codon mapping to residue "
        // + sourceDsPos + "(" + sourceChar + ")");
        // return;
        continue;
      }

      int mappedCodonStart = mappedPos[0]; // position (1...) of codon start
      int mappedCodonEnd = mappedPos[mappedPos.length - 1]; // codon end pos
      StringBuilder trailingCopiedGap = new StringBuilder();

      /*
       * Copy dna sequence up to and including this codon. Optionally, include
       * gaps before the codon starts (in introns) and/or after the codon starts
       * (in exons).
       * 
       * Note this only works for 'linear' splicing, not reverse or interleaved.
       * But then 'align dna as protein' doesn't make much sense otherwise.
       */
      int intronLength = 0;
      while (basesWritten + toOffset < mappedCodonEnd
              && thisSeqPos < thisSeq.length)
      {
        final char c = thisSeq[thisSeqPos++];
        if (c != myGapChar)
        {
          basesWritten++;
          int sourcePosition = basesWritten + toOffset;
          if (sourcePosition < mappedCodonStart)
          {
            /*
             * Found an unmapped (intron) base. First add in any preceding gaps
             * (if wanted).
             */
            if (preserveUnmappedGaps && trailingCopiedGap.length() > 0)
            {
              thisAligned.append(trailingCopiedGap.toString());
              intronLength += trailingCopiedGap.length();
              trailingCopiedGap = new StringBuilder();
            }
            intronLength++;
            inExon = false;
          }
          else
          {
            final boolean startOfCodon = sourcePosition == mappedCodonStart;
            int gapsToAdd = calculateGapsToInsert(preserveMappedGaps,
                    preserveUnmappedGaps, sourceGapMappedLength, inExon,
                    trailingCopiedGap.length(), intronLength, startOfCodon);
            for (int i = 0; i < gapsToAdd; i++)
            {
              thisAligned.append(myGapChar);
            }
            sourceGapMappedLength = 0;
            inExon = true;
          }
          thisAligned.append(c);
          trailingCopiedGap = new StringBuilder();
        }
        else
        {
          if (inExon && preserveMappedGaps)
          {
            trailingCopiedGap.append(myGapChar);
          }
          else if (!inExon && preserveUnmappedGaps)
          {
            trailingCopiedGap.append(myGapChar);
          }
        }
      }
    }

    /*
     * At end of model aligned sequence. Copy any remaining target sequence, optionally
     * including (intron) gaps.
     */
    while (thisSeqPos < thisSeq.length)
    {
      final char c = thisSeq[thisSeqPos++];
      if (c != myGapChar || preserveUnmappedGaps)
      {
        thisAligned.append(c);
      }
      sourceGapMappedLength--;
    }

    /*
     * finally add gaps to pad for any trailing source gaps or
     * unmapped characters
     */
    if (preserveUnmappedGaps)
    {
      while (sourceGapMappedLength > 0)
      {
        thisAligned.append(myGapChar);
        sourceGapMappedLength--;
      }
    }

    /*
     * All done aligning, set the aligned sequence.
     */
    alignTo.setSequence(new String(thisAligned));
  }

  /**
   * Helper method to work out how many gaps to insert when realigning.
   * 
   * @param preserveMappedGaps
   * @param preserveUnmappedGaps
   * @param sourceGapMappedLength
   * @param inExon
   * @param trailingCopiedGap
   * @param intronLength
   * @param startOfCodon
   * @return
   */
  protected static int calculateGapsToInsert(boolean preserveMappedGaps,
          boolean preserveUnmappedGaps, int sourceGapMappedLength,
          boolean inExon, int trailingGapLength, int intronLength,
          final boolean startOfCodon)
  {
    int gapsToAdd = 0;
    if (startOfCodon)
    {
      /*
       * Reached start of codon. Ignore trailing gaps in intron unless we are
       * preserving gaps in both exon and intron. Ignore them anyway if the
       * protein alignment introduces a gap at least as large as the intronic
       * region.
       */
      if (inExon && !preserveMappedGaps)
      {
        trailingGapLength = 0;
      }
      if (!inExon && !(preserveMappedGaps && preserveUnmappedGaps))
      {
        trailingGapLength = 0;
      }
      if (inExon)
      {
        gapsToAdd = Math.max(sourceGapMappedLength, trailingGapLength);
      }
      else
      {
        if (intronLength + trailingGapLength <= sourceGapMappedLength)
        {
          gapsToAdd = sourceGapMappedLength - intronLength;
        }
        else
        {
          gapsToAdd = Math.min(intronLength + trailingGapLength
                  - sourceGapMappedLength, trailingGapLength);
        }
      }
    }
    else
    {
      /*
       * second or third base of codon; check for any gaps in dna
       */
      if (!preserveMappedGaps)
      {
        trailingGapLength = 0;
      }
      gapsToAdd = Math.max(sourceGapMappedLength, trailingGapLength);
    }
    return gapsToAdd;
  }

  /**
   * Realigns the given protein to match the alignment of the dna, using codon
   * mappings to translate aligned codon positions to protein residues.
   * 
   * @param protein
   *          the alignment whose sequences are realigned by this method
   * @param dna
   *          the dna alignment whose alignment we are 'copying'
   * @return the number of sequences that were realigned
   */
  public static int alignProteinAsDna(AlignmentI protein, AlignmentI dna)
  {
    List<SequenceI> unmappedProtein = new ArrayList<SequenceI>();
    Map<AlignedCodon, Map<SequenceI, AlignedCodon>> alignedCodons = buildCodonColumnsMap(
            protein, dna, unmappedProtein);
    return alignProteinAs(protein, alignedCodons, unmappedProtein);
  }

  /**
   * Builds a map whose key is an aligned codon position (3 alignment column
   * numbers base 0), and whose value is a map from protein sequence to each
   * protein's peptide residue for that codon. The map generates an ordering of
   * the codons, and allows us to read off the peptides at each position in
   * order to assemble 'aligned' protein sequences.
   * 
   * @param protein
   *          the protein alignment
   * @param dna
   *          the coding dna alignment
   * @param unmappedProtein
   *          any unmapped proteins are added to this list
   * @return
   */
  protected static Map<AlignedCodon, Map<SequenceI, AlignedCodon>> buildCodonColumnsMap(
          AlignmentI protein, AlignmentI dna,
          List<SequenceI> unmappedProtein)
  {
    /*
     * maintain a list of any proteins with no mappings - these will be
     * rendered 'as is' in the protein alignment as we can't align them
     */
    unmappedProtein.addAll(protein.getSequences());

    List<AlignedCodonFrame> mappings = protein.getCodonFrames();

    /*
     * Map will hold, for each aligned codon position e.g. [3, 5, 6], a map of
     * {dnaSequence, {proteinSequence, codonProduct}} at that position. The
     * comparator keeps the codon positions ordered.
     */
    Map<AlignedCodon, Map<SequenceI, AlignedCodon>> alignedCodons = new TreeMap<AlignedCodon, Map<SequenceI, AlignedCodon>>(
            new CodonComparator());

    for (SequenceI dnaSeq : dna.getSequences())
    {
      for (AlignedCodonFrame mapping : mappings)
      {
        SequenceI prot = mapping.findAlignedSequence(dnaSeq, protein);
        if (prot != null)
        {
          Mapping seqMap = mapping.getMappingForSequence(dnaSeq);
          addCodonPositions(dnaSeq, prot, protein.getGapCharacter(),
                  seqMap, alignedCodons);
          unmappedProtein.remove(prot);
        }
      }
    }

    /*
     * Finally add any unmapped peptide start residues (e.g. for incomplete
     * codons) as if at the codon position before the second residue
     */
    // TODO resolve JAL-2022 so this fudge can be removed
    int mappedSequenceCount = protein.getHeight() - unmappedProtein.size();
    addUnmappedPeptideStarts(alignedCodons, mappedSequenceCount);
    
    return alignedCodons;
  }

  /**
   * Scans for any protein mapped from position 2 (meaning unmapped start
   * position e.g. an incomplete codon), and synthesizes a 'codon' for it at the
   * preceding position in the alignment
   * 
   * @param alignedCodons
   *          the codon-to-peptide map
   * @param mappedSequenceCount
   *          the number of distinct sequences in the map
   */
  protected static void addUnmappedPeptideStarts(
          Map<AlignedCodon, Map<SequenceI, AlignedCodon>> alignedCodons,
          int mappedSequenceCount)
  {
    // TODO delete this ugly hack once JAL-2022 is resolved
    // i.e. we can model startPhase > 0 (incomplete start codon)

    List<SequenceI> sequencesChecked = new ArrayList<SequenceI>();
    AlignedCodon lastCodon = null;
    Map<SequenceI, AlignedCodon> toAdd = new HashMap<SequenceI, AlignedCodon>();

    for (Entry<AlignedCodon, Map<SequenceI, AlignedCodon>> entry : alignedCodons
            .entrySet())
    {
      for (Entry<SequenceI, AlignedCodon> sequenceCodon : entry.getValue()
              .entrySet())
      {
        SequenceI seq = sequenceCodon.getKey();
        if (sequencesChecked.contains(seq))
        {
          continue;
        }
        sequencesChecked.add(seq);
        AlignedCodon codon = sequenceCodon.getValue();
        if (codon.peptideCol > 1)
        {
          System.err
                  .println("Problem mapping protein with >1 unmapped start positions: "
                          + seq.getName());
        }
        else if (codon.peptideCol == 1)
        {
          /*
           * first position (peptideCol == 0) was unmapped - add it
           */
          if (lastCodon != null)
          {
            AlignedCodon firstPeptide = new AlignedCodon(lastCodon.pos1,
                    lastCodon.pos2, lastCodon.pos3, String.valueOf(seq
                            .getCharAt(0)), 0);
            toAdd.put(seq, firstPeptide);
          }
          else
          {
            /*
             * unmapped residue at start of alignment (no prior column) -
             * 'insert' at nominal codon [0, 0, 0]
             */
            AlignedCodon firstPeptide = new AlignedCodon(0, 0, 0,
                    String.valueOf(seq.getCharAt(0)), 0);
            toAdd.put(seq, firstPeptide);
          }
        }
        if (sequencesChecked.size() == mappedSequenceCount)
        {
          // no need to check past first mapped position in all sequences
          break;
        }
      }
      lastCodon = entry.getKey();
    }

    /*
     * add any new codons safely after iterating over the map
     */
    for (Entry<SequenceI, AlignedCodon> startCodon : toAdd.entrySet())
    {
      addCodonToMap(alignedCodons, startCodon.getValue(),
              startCodon.getKey());
    }
  }

  /**
   * Update the aligned protein sequences to match the codon alignments given in
   * the map.
   * 
   * @param protein
   * @param alignedCodons
   *          an ordered map of codon positions (columns), with sequence/peptide
   *          values present in each column
   * @param unmappedProtein
   * @return
   */
  protected static int alignProteinAs(AlignmentI protein,
          Map<AlignedCodon, Map<SequenceI, AlignedCodon>> alignedCodons,
          List<SequenceI> unmappedProtein)
  {
    /*
     * Prefill aligned sequences with gaps before inserting aligned protein
     * residues.
     */
    int alignedWidth = alignedCodons.size();
    char[] gaps = new char[alignedWidth];
    Arrays.fill(gaps, protein.getGapCharacter());
    String allGaps = String.valueOf(gaps);
    for (SequenceI seq : protein.getSequences())
    {
      if (!unmappedProtein.contains(seq))
      {
        seq.setSequence(allGaps);
      }
    }

    int column = 0;
    for (AlignedCodon codon : alignedCodons.keySet())
    {
      final Map<SequenceI, AlignedCodon> columnResidues = alignedCodons
              .get(codon);
      for (Entry<SequenceI, AlignedCodon> entry : columnResidues.entrySet())
      {
        // place translated codon at its column position in sequence
        entry.getKey().getSequence()[column] = entry.getValue().product
                .charAt(0);
      }
      column++;
    }
    return 0;
  }

  /**
   * Populate the map of aligned codons by traversing the given sequence
   * mapping, locating the aligned positions of mapped codons, and adding those
   * positions and their translation products to the map.
   * 
   * @param dna
   *          the aligned sequence we are mapping from
   * @param protein
   *          the sequence to be aligned to the codons
   * @param gapChar
   *          the gap character in the dna sequence
   * @param seqMap
   *          a mapping to a sequence translation
   * @param alignedCodons
   *          the map we are building up
   */
  static void addCodonPositions(SequenceI dna, SequenceI protein,
          char gapChar, Mapping seqMap,
          Map<AlignedCodon, Map<SequenceI, AlignedCodon>> alignedCodons)
  {
    Iterator<AlignedCodon> codons = seqMap.getCodonIterator(dna, gapChar);

    /*
     * add codon positions, and their peptide translations, to the alignment
     * map, while remembering the first codon mapped
     */
    while (codons.hasNext())
    {
      try
      {
        AlignedCodon codon = codons.next();
        addCodonToMap(alignedCodons, codon, protein);
      } catch (IncompleteCodonException e)
      {
        // possible incomplete trailing codon - ignore
      } catch (NoSuchElementException e)
      {
        // possibly peptide lacking STOP
      }
    }
  }

  /**
   * Helper method to add a codon-to-peptide entry to the aligned codons map
   * 
   * @param alignedCodons
   * @param codon
   * @param protein
   */
  protected static void addCodonToMap(
          Map<AlignedCodon, Map<SequenceI, AlignedCodon>> alignedCodons,
          AlignedCodon codon, SequenceI protein)
  {
    Map<SequenceI, AlignedCodon> seqProduct = alignedCodons.get(codon);
    if (seqProduct == null)
    {
      seqProduct = new HashMap<SequenceI, AlignedCodon>();
      alignedCodons.put(codon, seqProduct);
    }
    seqProduct.put(protein, codon);
  }

  /**
   * Returns true if a cDNA/Protein mapping either exists, or could be made,
   * between at least one pair of sequences in the two alignments. Currently,
   * the logic is:
   * <ul>
   * <li>One alignment must be nucleotide, and the other protein</li>
   * <li>At least one pair of sequences must be already mapped, or mappable</li>
   * <li>Mappable means the nucleotide translation matches the protein sequence</li>
   * <li>The translation may ignore start and stop codons if present in the
   * nucleotide</li>
   * </ul>
   * 
   * @param al1
   * @param al2
   * @return
   */
  public static boolean isMappable(AlignmentI al1, AlignmentI al2)
  {
    if (al1 == null || al2 == null)
    {
      return false;
    }

    /*
     * Require one nucleotide and one protein
     */
    if (al1.isNucleotide() == al2.isNucleotide())
    {
      return false;
    }
    AlignmentI dna = al1.isNucleotide() ? al1 : al2;
    AlignmentI protein = dna == al1 ? al2 : al1;
    List<AlignedCodonFrame> mappings = protein.getCodonFrames();
    for (SequenceI dnaSeq : dna.getSequences())
    {
      for (SequenceI proteinSeq : protein.getSequences())
      {
        if (isMappable(dnaSeq, proteinSeq, mappings))
        {
          return true;
        }
      }
    }
    return false;
  }

  /**
   * Returns true if the dna sequence is mapped, or could be mapped, to the
   * protein sequence.
   * 
   * @param dnaSeq
   * @param proteinSeq
   * @param mappings
   * @return
   */
  protected static boolean isMappable(SequenceI dnaSeq,
          SequenceI proteinSeq, List<AlignedCodonFrame> mappings)
  {
    if (dnaSeq == null || proteinSeq == null)
    {
      return false;
    }

    SequenceI dnaDs = dnaSeq.getDatasetSequence() == null ? dnaSeq : dnaSeq
            .getDatasetSequence();
    SequenceI proteinDs = proteinSeq.getDatasetSequence() == null ? proteinSeq
            : proteinSeq.getDatasetSequence();

    for (AlignedCodonFrame mapping : mappings)
    {
      if (proteinDs == mapping.getAaForDnaSeq(dnaDs))
      {
        /*
         * already mapped
         */
        return true;
      }
    }

    /*
     * Just try to make a mapping (it is not yet stored), test whether
     * successful.
     */
    return mapCdnaToProtein(proteinDs, dnaDs) != null;
  }

  /**
   * Finds any reference annotations associated with the sequences in
   * sequenceScope, that are not already added to the alignment, and adds them
   * to the 'candidates' map. Also populates a lookup table of annotation
   * labels, keyed by calcId, for use in constructing tooltips or the like.
   * 
   * @param sequenceScope
   *          the sequences to scan for reference annotations
   * @param labelForCalcId
   *          (optional) map to populate with label for calcId
   * @param candidates
   *          map to populate with annotations for sequence
   * @param al
   *          the alignment to check for presence of annotations
   */
  public static void findAddableReferenceAnnotations(
          List<SequenceI> sequenceScope,
          Map<String, String> labelForCalcId,
          final Map<SequenceI, List<AlignmentAnnotation>> candidates,
          AlignmentI al)
  {
    if (sequenceScope == null)
    {
      return;
    }

    /*
     * For each sequence in scope, make a list of any annotations on the
     * underlying dataset sequence which are not already on the alignment.
     * 
     * Add to a map of { alignmentSequence, <List of annotations to add> }
     */
    for (SequenceI seq : sequenceScope)
    {
      SequenceI dataset = seq.getDatasetSequence();
      if (dataset == null)
      {
        continue;
      }
      AlignmentAnnotation[] datasetAnnotations = dataset.getAnnotation();
      if (datasetAnnotations == null)
      {
        continue;
      }
      final List<AlignmentAnnotation> result = new ArrayList<AlignmentAnnotation>();
      for (AlignmentAnnotation dsann : datasetAnnotations)
      {
        /*
         * Find matching annotations on the alignment. If none is found, then
         * add this annotation to the list of 'addable' annotations for this
         * sequence.
         */
        final Iterable<AlignmentAnnotation> matchedAlignmentAnnotations = al
                .findAnnotations(seq, dsann.getCalcId(), dsann.label);
        if (!matchedAlignmentAnnotations.iterator().hasNext())
        {
          result.add(dsann);
          if (labelForCalcId != null)
          {
            labelForCalcId.put(dsann.getCalcId(), dsann.label);
          }
        }
      }
      /*
       * Save any addable annotations for this sequence
       */
      if (!result.isEmpty())
      {
        candidates.put(seq, result);
      }
    }
  }

  /**
   * Adds annotations to the top of the alignment annotations, in the same order
   * as their related sequences.
   * 
   * @param annotations
   *          the annotations to add
   * @param alignment
   *          the alignment to add them to
   * @param selectionGroup
   *          current selection group (or null if none)
   */
  public static void addReferenceAnnotations(
          Map<SequenceI, List<AlignmentAnnotation>> annotations,
          final AlignmentI alignment, final SequenceGroup selectionGroup)
  {
    for (SequenceI seq : annotations.keySet())
    {
      for (AlignmentAnnotation ann : annotations.get(seq))
      {
        AlignmentAnnotation copyAnn = new AlignmentAnnotation(ann);
        int startRes = 0;
        int endRes = ann.annotations.length;
        if (selectionGroup != null)
        {
          startRes = selectionGroup.getStartRes();
          endRes = selectionGroup.getEndRes();
        }
        copyAnn.restrict(startRes, endRes);

        /*
         * Add to the sequence (sets copyAnn.datasetSequence), unless the
         * original annotation is already on the sequence.
         */
        if (!seq.hasAnnotation(ann))
        {
          seq.addAlignmentAnnotation(copyAnn);
        }
        // adjust for gaps
        copyAnn.adjustForAlignment();
        // add to the alignment and set visible
        alignment.addAnnotation(copyAnn);
        copyAnn.visible = true;
      }
    }
  }

  /**
   * Set visibility of alignment annotations of specified types (labels), for
   * specified sequences. This supports controls like
   * "Show all secondary structure", "Hide all Temp factor", etc.
   * 
   * @al the alignment to scan for annotations
   * @param types
   *          the types (labels) of annotations to be updated
   * @param forSequences
   *          if not null, only annotations linked to one of these sequences are
   *          in scope for update; if null, acts on all sequence annotations
   * @param anyType
   *          if this flag is true, 'types' is ignored (label not checked)
   * @param doShow
   *          if true, set visibility on, else set off
   */
  public static void showOrHideSequenceAnnotations(AlignmentI al,
          Collection<String> types, List<SequenceI> forSequences,
          boolean anyType, boolean doShow)
  {
    for (AlignmentAnnotation aa : al.getAlignmentAnnotation())
    {
      if (anyType || types.contains(aa.label))
      {
        if ((aa.sequenceRef != null)
                && (forSequences == null || forSequences
                        .contains(aa.sequenceRef)))
        {
          aa.visible = doShow;
        }
      }
    }
  }

  /**
   * Returns true if either sequence has a cross-reference to the other
   * 
   * @param seq1
   * @param seq2
   * @return
   */
  public static boolean haveCrossRef(SequenceI seq1, SequenceI seq2)
  {
    // Note: moved here from class CrossRef as the latter class has dependencies
    // not availability to the applet's classpath
    return hasCrossRef(seq1, seq2) || hasCrossRef(seq2, seq1);
  }

  /**
   * Returns true if seq1 has a cross-reference to seq2. Currently this assumes
   * that sequence name is structured as Source|AccessionId.
   * 
   * @param seq1
   * @param seq2
   * @return
   */
  public static boolean hasCrossRef(SequenceI seq1, SequenceI seq2)
  {
    if (seq1 == null || seq2 == null)
    {
      return false;
    }
    String name = seq2.getName();
    final DBRefEntry[] xrefs = seq1.getDBRefs();
    if (xrefs != null)
    {
      for (DBRefEntry xref : xrefs)
      {
        String xrefName = xref.getSource() + "|" + xref.getAccessionId();
        // case-insensitive test, consistent with DBRefEntry.equalRef()
        if (xrefName.equalsIgnoreCase(name))
        {
          return true;
        }
      }
    }
    return false;
  }

  /**
   * Constructs an alignment consisting of the mapped (CDS) regions in the given
   * nucleotide sequences, and updates mappings to match. The CDS sequences are
   * added to the original alignment's dataset, which is shared by the new
   * alignment. Mappings from nucleotide to CDS, and from CDS to protein, are
   * added to the alignment dataset.
   * 
   * @param dna
   *          aligned dna sequences
   * @param mappings
   *          from dna to protein
   * @param al
   * @return an alignment whose sequences are the cds-only parts of the dna
   *         sequences (or null if no mappings are found)
   */
  public static AlignmentI makeCdsAlignment(SequenceI[] dna,
          List<AlignedCodonFrame> mappings, AlignmentI al)
  {
    List<SequenceI> cdsSeqs = new ArrayList<SequenceI>();
    
    for (SequenceI seq : dna)
    {
      AlignedCodonFrame cdsMappings = new AlignedCodonFrame();
      List<AlignedCodonFrame> seqMappings = MappingUtils
              .findMappingsForSequence(seq, mappings);
      List<AlignedCodonFrame> alignmentMappings = al.getCodonFrames();
      for (AlignedCodonFrame mapping : seqMappings)
      {
        for (Mapping aMapping : mapping.getMappingsFromSequence(seq))
        {
          SequenceI cdsSeq = makeCdsSequence(seq.getDatasetSequence(),
                  aMapping);
          cdsSeqs.add(cdsSeq);
    
          /*
           * add a mapping from CDS to the (unchanged) mapped to range
           */
          List<int[]> cdsRange = Collections.singletonList(new int[] { 1,
              cdsSeq.getLength() });
          MapList map = new MapList(cdsRange, aMapping.getMap()
                  .getToRanges(), aMapping.getMap().getFromRatio(),
                  aMapping.getMap().getToRatio());
          cdsMappings.addMap(cdsSeq, aMapping.getTo(), map);

          /*
           * add another mapping from original 'from' range to CDS
           */
          map = new MapList(aMapping.getMap().getFromRanges(), cdsRange, 1,
                  1);
          cdsMappings.addMap(seq.getDatasetSequence(), cdsSeq, map);

          alignmentMappings.add(cdsMappings);

          /*
           * transfer any features on dna that overlap the CDS
           */
          transferFeatures(seq, cdsSeq, map, null, SequenceOntologyI.CDS);
        }
      }
    }

    /*
     * add CDS seqs to shared dataset
     */
    Alignment dataset = al.getDataset();
    for (SequenceI seq : cdsSeqs)
    {
      if (!dataset.getSequences().contains(seq.getDatasetSequence()))
      {
        dataset.addSequence(seq.getDatasetSequence());
      }
    }
    AlignmentI cds = new Alignment(cdsSeqs.toArray(new SequenceI[cdsSeqs
            .size()]));
    cds.setDataset(dataset);

    return cds;
  }

  /**
   * Helper method that makes a CDS sequence as defined by the mappings from the
   * given sequence i.e. extracts the 'mapped from' ranges (which may be on
   * forward or reverse strand).
   * 
   * @param seq
   * @param mapping
   * @return
   */
  static SequenceI makeCdsSequence(SequenceI seq, Mapping mapping)
  {
    char[] seqChars = seq.getSequence();
    List<int[]> fromRanges = mapping.getMap().getFromRanges();
    int cdsWidth = MappingUtils.getLength(fromRanges);
    char[] newSeqChars = new char[cdsWidth];

    int newPos = 0;
    for (int[] range : fromRanges)
    {
      if (range[0] <= range[1])
      {
        // forward strand mapping - just copy the range
        int length = range[1] - range[0] + 1;
        System.arraycopy(seqChars, range[0] - 1, newSeqChars, newPos,
                length);
        newPos += length;
      }
      else
      {
        // reverse strand mapping - copy and complement one by one
        for (int i = range[0]; i >= range[1]; i--)
        {
          newSeqChars[newPos++] = Dna.getComplement(seqChars[i - 1]);
        }
      }
    }

    SequenceI newSeq = new Sequence(seq.getName() + "|"
            + mapping.getTo().getName(), newSeqChars, 1, newPos);
    newSeq.createDatasetSequence();
    return newSeq;
  }

  /**
   * Transfers co-located features on 'fromSeq' to 'toSeq', adjusting the
   * feature start/end ranges, optionally omitting specified feature types.
   * Returns the number of features copied.
   * 
   * @param fromSeq
   * @param toSeq
   * @param select
   *          if not null, only features of this type are copied (including
   *          subtypes in the Sequence Ontology)
   * @param mapping
   *          the mapping from 'fromSeq' to 'toSeq'
   * @param omitting
   */
  public static int transferFeatures(SequenceI fromSeq, SequenceI toSeq,
          MapList mapping, String select, String... omitting)
  {
    SequenceI copyTo = toSeq;
    while (copyTo.getDatasetSequence() != null)
    {
      copyTo = copyTo.getDatasetSequence();
    }

    SequenceOntologyI so = SequenceOntologyFactory.getInstance();
    int count = 0;
    SequenceFeature[] sfs = fromSeq.getSequenceFeatures();
    if (sfs != null)
    {
      for (SequenceFeature sf : sfs)
      {
        String type = sf.getType();
        if (select != null && !so.isA(type, select))
        {
          continue;
        }
        boolean omit = false;
        for (String toOmit : omitting)
        {
          if (type.equals(toOmit))
          {
            omit = true;
          }
        }
        if (omit)
        {
          continue;
        }

        /*
         * locate the mapped range - null if either start or end is
         * not mapped (no partial overlaps are calculated)
         */
        int start = sf.getBegin();
        int end = sf.getEnd();
        int[] mappedTo = mapping.locateInTo(start, end);
        /*
         * if whole exon range doesn't map, try interpreting it
         * as 5' or 3' exon overlapping the CDS range
         */
        if (mappedTo == null)
        {
          mappedTo = mapping.locateInTo(end, end);
          if (mappedTo != null)
          {
            /*
             * end of exon is in CDS range - 5' overlap
             * to a range from the start of the peptide
             */
            mappedTo[0] = 1;
          }
        }
        if (mappedTo == null)
        {
          mappedTo = mapping.locateInTo(start, start);
          if (mappedTo != null)
          {
            /*
             * start of exon is in CDS range - 3' overlap
             * to a range up to the end of the peptide
             */
            mappedTo[1] = toSeq.getLength();
          }
        }
        if (mappedTo != null)
        {
          SequenceFeature copy = new SequenceFeature(sf);
          copy.setBegin(Math.min(mappedTo[0], mappedTo[1]));
          copy.setEnd(Math.max(mappedTo[0], mappedTo[1]));
          copyTo.addSequenceFeature(copy);
          count++;
        }
      }
    }
    return count;
  }

  /**
   * Returns a mapping from dna to protein by inspecting sequence features of
   * type "CDS" on the dna.
   * 
   * @param dnaSeq
   * @param proteinSeq
   * @return
   */
  public static MapList mapCdsToProtein(SequenceI dnaSeq,
          SequenceI proteinSeq)
  {
    List<int[]> ranges = findCdsPositions(dnaSeq);
    int mappedDnaLength = MappingUtils.getLength(ranges);

    int proteinLength = proteinSeq.getLength();
    int proteinStart = proteinSeq.getStart();
    int proteinEnd = proteinSeq.getEnd();

    /*
     * incomplete start codon may mean X at start of peptide
     * we ignore both for mapping purposes
     */
    if (proteinSeq.getCharAt(0) == 'X')
    {
      // todo JAL-2022 support startPhase > 0
      proteinStart++;
      proteinLength--;
    }
    List<int[]> proteinRange = new ArrayList<int[]>();

    /*
     * dna length should map to protein (or protein plus stop codon)
     */
    int codesForResidues = mappedDnaLength / 3;
    if (codesForResidues == (proteinLength + 1))
    {
      // assuming extra codon is for STOP and not in peptide
      codesForResidues--;
    }
    if (codesForResidues == proteinLength)
    {
      proteinRange.add(new int[] { proteinStart, proteinEnd });
      return new MapList(ranges, proteinRange, 3, 1);
    }
    return null;
  }

  /**
   * Returns a list of CDS ranges found (as sequence positions base 1), i.e. of
   * start/end positions of sequence features of type "CDS" (or a sub-type of
   * CDS in the Sequence Ontology). The ranges are sorted into ascending start
   * position order, so this method is only valid for linear CDS in the same
   * sense as the protein product.
   * 
   * @param dnaSeq
   * @return
   */
  public static List<int[]> findCdsPositions(SequenceI dnaSeq)
  {
    List<int[]> result = new ArrayList<int[]>();
    SequenceFeature[] sfs = dnaSeq.getSequenceFeatures();
    if (sfs == null)
    {
      return result;
    }

    SequenceOntologyI so = SequenceOntologyFactory.getInstance();
    int startPhase = 0;

    for (SequenceFeature sf : sfs)
    {
      /*
       * process a CDS feature (or a sub-type of CDS)
       */
      if (so.isA(sf.getType(), SequenceOntologyI.CDS))
      {
        int phase = 0;
        try
        {
          phase = Integer.parseInt(sf.getPhase());
        } catch (NumberFormatException e)
        {
          // ignore
        }
        /*
         * phase > 0 on first codon means 5' incomplete - skip to the start
         * of the next codon; example ENST00000496384
         */
        int begin = sf.getBegin();
        int end = sf.getEnd();
        if (result.isEmpty())
        {
          begin += phase;
          if (begin > end)
          {
            // shouldn't happen!
            System.err
                    .println("Error: start phase extends beyond start CDS in "
                            + dnaSeq.getName());
          }
        }
        result.add(new int[] { begin, end });
      }
    }

    /*
     * remove 'startPhase' positions (usually 0) from the first range 
     * so we begin at the start of a complete codon
     */
    if (!result.isEmpty())
    {
      // TODO JAL-2022 correctly model start phase > 0
      result.get(0)[0] += startPhase;
    }

    /*
     * Finally sort ranges by start position. This avoids a dependency on 
     * keeping features in order on the sequence (if they are in order anyway,
     * the sort will have almost no work to do). The implicit assumption is CDS
     * ranges are assembled in order. Other cases should not use this method,
     * but instead construct an explicit mapping for CDS (e.g. EMBL parsing).
     */
    Collections.sort(result, new Comparator<int[]>()
    {
      @Override
      public int compare(int[] o1, int[] o2)
      {
        return Integer.compare(o1[0], o2[0]);
      }
    });
    return result;
  }

  /**
   * Maps exon features from dna to protein, and computes variants in peptide
   * product generated by variants in dna, and adds them as sequence_variant
   * features on the protein sequence. Returns the number of variant features
   * added.
   * 
   * @param dnaSeq
   * @param peptide
   * @param dnaToProtein
   */
  public static int computeProteinFeatures(SequenceI dnaSeq,
          SequenceI peptide, MapList dnaToProtein)
  {
    while (dnaSeq.getDatasetSequence() != null)
    {
      dnaSeq = dnaSeq.getDatasetSequence();
    }
    while (peptide.getDatasetSequence() != null)
    {
      peptide = peptide.getDatasetSequence();
    }

    transferFeatures(dnaSeq, peptide, dnaToProtein, SequenceOntologyI.EXON);

    /*
     * compute protein variants from dna variants and codon mappings;
     * NB - alternatively we could retrieve this using the REST service e.g.
     * http://rest.ensembl.org/overlap/translation
     * /ENSP00000288602?feature=transcript_variation;content-type=text/xml
     * which would be a bit slower but possibly more reliable
     */
    LinkedHashMap<Integer, String[][]> variants = buildDnaVariantsMap(
            dnaSeq, dnaToProtein);

    /*
     * scan codon variations, compute peptide variants and add to peptide sequence
     */
    int count = 0;
    for (Entry<Integer, String[][]> variant : variants.entrySet())
    {
      int peptidePos = variant.getKey();
      String[][] codonVariants = variant.getValue();
      String residue = String.valueOf(peptide.getCharAt(peptidePos - 1)); // 0-based
      List<String> peptideVariants = computePeptideVariants(codonVariants,
              residue);
      if (!peptideVariants.isEmpty())
      {
        String desc = residue + "," // include canonical residue in description
                + StringUtils.listToDelimitedString(peptideVariants, ", ");
        SequenceFeature sf = new SequenceFeature(
                SequenceOntologyI.SEQUENCE_VARIANT, desc, peptidePos,
                peptidePos, 0f, null);
        peptide.addSequenceFeature(sf);
        count++;
      }
    }

    /*
     * ugly sort to get sequence features in start position order
     * - would be better to store in Sequence as a TreeSet instead?
     */
    Arrays.sort(peptide.getSequenceFeatures(),
            new Comparator<SequenceFeature>()
            {
              @Override
              public int compare(SequenceFeature o1, SequenceFeature o2)
              {
                int c = Integer.compare(o1.getBegin(), o2.getBegin());
                return c == 0 ? Integer.compare(o1.getEnd(), o2.getEnd())
                        : c;
              }
            });
    return count;
  }

  /**
   * Builds a map whose key is position in the protein sequence, and value is an
   * array of all variants for the coding codon positions
   * 
   * @param dnaSeq
   * @param dnaToProtein
   * @return
   */
  static LinkedHashMap<Integer, String[][]> buildDnaVariantsMap(
          SequenceI dnaSeq, MapList dnaToProtein)
  {
    /*
     * map from peptide position to all variant features of the codon for it
     * LinkedHashMap ensures we add the peptide features in sequence order
     */
    LinkedHashMap<Integer, String[][]> variants = new LinkedHashMap<Integer, String[][]>();
    SequenceOntologyI so = SequenceOntologyFactory.getInstance();

    SequenceFeature[] dnaFeatures = dnaSeq.getSequenceFeatures();
    if (dnaFeatures == null)
    {
      return variants;
    }

    int dnaStart = dnaSeq.getStart();
    int[] lastCodon = null;
    int lastPeptidePostion = 0;

    /*
     * build a map of codon variations for peptides
     */
    for (SequenceFeature sf : dnaFeatures)
    {
      int dnaCol = sf.getBegin();
      if (dnaCol != sf.getEnd())
      {
        // not handling multi-locus variant features
        continue;
      }
      if (so.isA(sf.getType(), SequenceOntologyI.SEQUENCE_VARIANT))
      {
        int[] mapsTo = dnaToProtein.locateInTo(dnaCol, dnaCol);
        if (mapsTo == null)
        {
          // feature doesn't lie within coding region
          continue;
        }
        int peptidePosition = mapsTo[0];
        String[][] codonVariants = variants.get(peptidePosition);
        if (codonVariants == null)
        {
          codonVariants = new String[3][];
          variants.put(peptidePosition, codonVariants);
        }

        /*
         * extract dna variants to a string array
         */
        String alls = (String) sf.getValue("alleles");
        if (alls == null)
        {
          continue;
        }
        String[] alleles = alls.toUpperCase().split(",");
        int i = 0;
        for (String allele : alleles)
        {
          alleles[i++] = allele.trim(); // lose any space characters "A, G"
        }

        /*
         * get this peptide's codon positions e.g. [3, 4, 5] or [4, 7, 10]
         */
        int[] codon = peptidePosition == lastPeptidePostion ? lastCodon
                : MappingUtils.flattenRanges(dnaToProtein.locateInFrom(
                        peptidePosition, peptidePosition));
        lastPeptidePostion = peptidePosition;
        lastCodon = codon;

        /*
         * save nucleotide (and this variant) for each codon position
         */
        for (int codonPos = 0; codonPos < 3; codonPos++)
        {
          String nucleotide = String.valueOf(
                  dnaSeq.getCharAt(codon[codonPos] - dnaStart))
                  .toUpperCase();
          if (codonVariants[codonPos] == null)
          {
            /*
             * record current dna base
             */
            codonVariants[codonPos] = new String[] { nucleotide };
          }
          if (codon[codonPos] == dnaCol)
          {
            /*
             * add alleles to dna base (and any previously found alleles)
             */
            String[] known = codonVariants[codonPos];
            String[] dnaVariants = new String[alleles.length + known.length];
            System.arraycopy(known, 0, dnaVariants, 0, known.length);
            System.arraycopy(alleles, 0, dnaVariants, known.length,
                    alleles.length);
            codonVariants[codonPos] = dnaVariants;
          }
        }
      }
    }
    return variants;
  }

  /**
   * Returns a sorted, non-redundant list of all peptide translations generated
   * by the given dna variants, excluding the current residue value
   * 
   * @param codonVariants
   *          an array of base values (acgtACGT) for codon positions 1, 2, 3
   * @param residue
   *          the current residue translation
   * @return
   */
  static List<String> computePeptideVariants(String[][] codonVariants,
          String residue)
  {
    List<String> result = new ArrayList<String>();
    for (String base1 : codonVariants[0])
    {
      for (String base2 : codonVariants[1])
      {
        for (String base3 : codonVariants[2])
        {
          String codon = base1 + base2 + base3;
          /*
           * get peptide translation of codon e.g. GAT -> D
           * note that variants which are not single alleles,
           * e.g. multibase variants or HGMD_MUTATION etc
           * are ignored here
           */
          String peptide = codon.contains("-") ? "-"
                  : (codon.length() > 3 ? null : ResidueProperties
                          .codonTranslate(codon));
          if (peptide != null && !result.contains(peptide)
                  && !peptide.equalsIgnoreCase(residue))
          {
            result.add(peptide);
          }
        }
      }
    }

    /*
     * sort alphabetically with STOP at the end
     */
    Collections.sort(result, new Comparator<String>()
    {

      @Override
      public int compare(String o1, String o2)
      {
        if ("STOP".equals(o1))
        {
          return 1;
        }
        else if ("STOP".equals(o2))
        {
          return -1;
        }
        else
        {
          return o1.compareTo(o2);
        }
      }
    });
    return result;
  }

  /**
   * Makes an alignment with a copy of the given sequences, adding in any
   * non-redundant sequences which are mapped to by the cross-referenced
   * sequences.
   * 
   * @param seqs
   * @param xrefs
   * @return
   */
  public static AlignmentI makeCopyAlignment(SequenceI[] seqs,
          SequenceI[] xrefs)
  {
    AlignmentI copy = new Alignment(new Alignment(seqs));

    SequenceIdMatcher matcher = new SequenceIdMatcher(seqs);
    if (xrefs != null)
    {
      for (SequenceI xref : xrefs)
      {
        DBRefEntry[] dbrefs = xref.getDBRefs();
        if (dbrefs != null)
        {
          for (DBRefEntry dbref : dbrefs)
          {
            if (dbref.getMap() == null || dbref.getMap().getTo() == null)
            {
              continue;
            }
            SequenceI mappedTo = dbref.getMap().getTo();
            SequenceI match = matcher.findIdMatch(mappedTo);
            if (match == null)
            {
              matcher.add(mappedTo);
              copy.addSequence(mappedTo);
            }
          }
        }
      }
    }
    return copy;
  }

  /**
   * Try to align sequences in 'unaligned' to match the alignment of their
   * mapped regions in 'aligned'. For example, could use this to align CDS
   * sequences which are mapped to their parent cDNA sequences.
   * 
   * This method handles 1:1 mappings (dna-to-dna or protein-to-protein). For
   * dna-to-protein or protein-to-dna use alternative methods.
   * 
   * @param unaligned
   *          sequences to be aligned
   * @param aligned
   *          holds aligned sequences and their mappings
   * @return
   */
  public static int alignAs(AlignmentI unaligned, AlignmentI aligned)
  {
    List<SequenceI> unmapped = new ArrayList<SequenceI>();
    Map<Integer, Map<SequenceI, Character>> columnMap = buildMappedColumnsMap(
            unaligned, aligned, unmapped);
    int width = columnMap.size();
    char gap = unaligned.getGapCharacter();
    int realignedCount = 0;

    for (SequenceI seq : unaligned.getSequences())
    {
      if (!unmapped.contains(seq))
      {
        char[] newSeq = new char[width];
        Arrays.fill(newSeq, gap);
        int newCol = 0;
        int lastCol = 0;

        /*
         * traverse the map to find columns populated
         * by our sequence
         */
        for (Integer column : columnMap.keySet())
        {
          Character c = columnMap.get(column).get(seq);
          if (c != null)
          {
            /*
             * sequence has a character at this position
             * 
             */
            newSeq[newCol] = c;
            lastCol = newCol;
          }
          newCol++;
        }
        
        /*
         * trim trailing gaps
         */
        if (lastCol < width)
        {
          char[] tmp = new char[lastCol + 1];
          System.arraycopy(newSeq, 0, tmp, 0, lastCol + 1);
          newSeq = tmp;
        }
        seq.setSequence(String.valueOf(newSeq));
        realignedCount++;
      }
    }
    return realignedCount;
  }

  /**
   * Returns a map whose key is alignment column number (base 1), and whose
   * values are a map of sequence characters in that column.
   * 
   * @param unaligned
   * @param aligned
   * @param unmapped
   * @return
   */
  static Map<Integer, Map<SequenceI, Character>> buildMappedColumnsMap(
          AlignmentI unaligned, AlignmentI aligned, List<SequenceI> unmapped)
  {
    /*
     * Map will hold, for each aligned column position, a map of
     * {unalignedSequence, sequenceCharacter} at that position.
     * TreeMap keeps the entries in ascending column order. 
     */
    Map<Integer, Map<SequenceI, Character>> map = new TreeMap<Integer, Map<SequenceI, Character>>();

    /*
     * r any sequences that have no mapping so can't be realigned
     */
    unmapped.addAll(unaligned.getSequences());

    List<AlignedCodonFrame> mappings = aligned.getCodonFrames();

    for (SequenceI seq : unaligned.getSequences())
    {
      for (AlignedCodonFrame mapping : mappings)
      {
        SequenceI fromSeq = mapping.findAlignedSequence(seq, aligned);
        if (fromSeq != null)
        {
          Mapping seqMap = mapping.getMappingBetween(fromSeq, seq);
          if (addMappedPositions(seq, fromSeq, seqMap, map))
          {
            unmapped.remove(seq);
          }
        }
      }
    }
    return map;
  }

  /**
   * Helper method that adds to a map the mapped column positions of a sequence. <br>
   * For example if aaTT-Tg-gAAA is mapped to TTTAAA then the map should record
   * that columns 3,4,6,10,11,12 map to characters T,T,T,A,A,A of the mapped to
   * sequence.
   * 
   * @param seq
   *          the sequence whose column positions we are recording
   * @param fromSeq
   *          a sequence that is mapped to the first sequence
   * @param seqMap
   *          the mapping from 'fromSeq' to 'seq'
   * @param map
   *          a map to add the column positions (in fromSeq) of the mapped
   *          positions of seq
   * @return
   */
  static boolean addMappedPositions(SequenceI seq, SequenceI fromSeq,
          Mapping seqMap, Map<Integer, Map<SequenceI, Character>> map)
  {
    if (seqMap == null)
    {
      return false;
    }

    char[] fromChars = fromSeq.getSequence();
    int toStart = seq.getStart();
    char[] toChars = seq.getSequence();

    /*
     * traverse [start, end, start, end...] ranges in fromSeq
     */
    for (int[] fromRange : seqMap.getMap().getFromRanges())
    {
      for (int i = 0; i < fromRange.length - 1; i += 2)
      {
        boolean forward = fromRange[i + 1] >= fromRange[i];

        /*
         * find the range mapped to (sequence positions base 1)
         */
        int[] range = seqMap.locateMappedRange(fromRange[i],
                fromRange[i + 1]);
        if (range == null)
        {
          System.err.println("Error in mapping " + seqMap + " from "
                  + fromSeq.getName());
          return false;
        }
        int fromCol = fromSeq.findIndex(fromRange[i]);
        int mappedCharPos = range[0];

        /*
         * walk over the 'from' aligned sequence in forward or reverse
         * direction; when a non-gap is found, record the column position
         * of the next character of the mapped-to sequence; stop when all
         * the characters of the range have been counted
         */
        while (mappedCharPos <= range[1])
        {
          if (!Comparison.isGap(fromChars[fromCol - 1]))
          {
            /*
             * mapped from sequence has a character in this column
             * record the column position for the mapped to character
             */
            Map<SequenceI, Character> seqsMap = map.get(fromCol);
            if (seqsMap == null)
            {
              seqsMap = new HashMap<SequenceI, Character>();
              map.put(fromCol, seqsMap);
            }
            seqsMap.put(seq, toChars[mappedCharPos - toStart]);
            mappedCharPos++;
          }
          fromCol += (forward ? 1 : -1);
        }
      }
    }
    return true;
  }

  // strictly temporary hack until proper criteria for aligning protein to cds
  // are in place; this is so Ensembl -> fetch xrefs Uniprot aligns the Uniprot
  public static boolean looksLikeEnsembl(AlignmentI alignment)
  {
    for (SequenceI seq : alignment.getSequences())
    {
      String name = seq.getName();
      if (!name.startsWith("ENSG") && !name.startsWith("ENST"))
      {
        return false;
      }
    }
    return true;
  }
}