/* * 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 . * The Jalview Authors are detailed in the 'AUTHORS' file. */ package MCview; import jalview.analysis.AlignSeq; import jalview.datamodel.AlignmentAnnotation; import jalview.datamodel.Annotation; import jalview.datamodel.Mapping; import jalview.datamodel.Sequence; import jalview.datamodel.SequenceFeature; import jalview.datamodel.SequenceI; import jalview.schemes.ColourSchemeI; import jalview.schemes.ResidueProperties; import jalview.structure.StructureMapping; import java.awt.Color; import java.util.List; import java.util.Vector; public class PDBChain { /** * SequenceFeature group for PDB File features added to sequences */ private static final String PDBFILEFEATURE = "PDBFile"; private static final String IEASTATUS = "IEA:jalview"; public String id; public Vector bonds = new Vector(); public Vector atoms = new Vector(); public Vector residues = new Vector(); public int offset; /** * sequence is the sequence extracted by the chain parsing code */ public SequenceI sequence; /** * shadow is the sequence created by any other parsing processes (e.g. Jmol, * RNAview) */ public SequenceI shadow = null; public boolean isNa = false; public boolean isVisible = true; public int pdbstart = 0; public int pdbend = 0; public int seqstart = 0; public int seqend = 0; public String pdbid = ""; public PDBChain(String pdbid, String id) { this.pdbid = pdbid.toLowerCase(); this.id = id; } /** * character used to write newlines */ protected String newline = System.getProperty("line.separator"); public Mapping shadowMap; public void setNewlineString(String nl) { newline = nl; } public String getNewlineString() { return newline; } public String print() { StringBuilder tmp = new StringBuilder(256); for (Bond b : bonds) { tmp.append(b.at1.resName).append(" ").append(b.at1.resNumber) .append(" ").append(offset).append(newline); } return tmp.toString(); } /** * Annotate the residues with their corresponding positions in s1 using the * alignment in as NOTE: This clears all atom.alignmentMapping values on the * structure. * * @param as * @param s1 */ public void makeExactMapping(AlignSeq as, SequenceI s1) { int pdbpos = as.getSeq2Start() - 2; int alignpos = s1.getStart() + as.getSeq1Start() - 3; // first clear out any old alignmentMapping values: for (Atom atom : atoms) { atom.alignmentMapping = -1; } // and now trace the alignment onto the atom set. for (int i = 0; i < as.astr1.length(); i++) { if (as.astr1.charAt(i) != '-') { alignpos++; } if (as.astr2.charAt(i) != '-') { pdbpos++; } if (as.astr1.charAt(i) == as.astr2.charAt(i)) { Residue res = residues.elementAt(pdbpos); for (Atom atom : res.atoms) { atom.alignmentMapping = alignpos; } } } } /** * copy over the RESNUM seqfeatures from the internal chain sequence to the * mapped sequence * * @param seq * @param status * The Status of the transferred annotation * @return the features added to sq (or its dataset) */ public SequenceFeature[] transferRESNUMFeatures(SequenceI seq, String status) { SequenceI sq = seq; while (sq != null && sq.getDatasetSequence() != null) { sq = sq.getDatasetSequence(); if (sq == sequence) { return null; } } /** * Remove any existing features for this chain if they exist ? * SequenceFeature[] seqsfeatures=seq.getSequenceFeatures(); int * totfeat=seqsfeatures.length; // Remove any features for this exact chain * ? for (int i=0; i 99% 'P', flag as nucleotide; note the count doesn't include the last * residue */ if (residues.size() > 0 && (numNa / (residues.size() - 1) > 0.99)) { isNa = true; } } /** * Construct a bond from atom1 to atom2 and add it to the list of bonds for * this chain * * @param at1 * @param at2 */ public void makeBond(Atom at1, Atom at2) { bonds.addElement(new Bond(at1, at2)); } /** * Traverses the list of atoms and *
    *
  • constructs a list of Residues, each containing all the atoms that share * the same residue number
  • *
  • adds a RESNUM sequence feature for each position
  • *
  • creates the sequence string
  • *
  • determines if nucleotide
  • *
  • saves the residue number of the first atom as 'offset'
  • *
  • adds temp factor annotation if the flag is set to do so
  • *
* * @param visibleChainAnnotation */ public void makeResidueList(boolean visibleChainAnnotation) { int count = 0; Object symbol; boolean deoxyn = false; boolean nucleotide = false; StringBuilder seq = new StringBuilder(256); Vector resFeatures = new Vector(); Vector resAnnotation = new Vector(); int i, iSize = atoms.size() - 1; int resNumber = -1; for (i = 0; i <= iSize; i++) { Atom tmp = atoms.elementAt(i); resNumber = tmp.resNumber; int res = resNumber; if (i == 0) { offset = resNumber; } Vector resAtoms = new Vector(); // Add atoms to a vector while the residue number // remains the same as the first atom's resNumber (res) while ((resNumber == res) && (i < atoms.size())) { resAtoms.add(atoms.elementAt(i)); i++; if (i < atoms.size()) { resNumber = atoms.elementAt(i).resNumber; } else { resNumber++; } } // We need this to keep in step with the outer for i = loop i--; // Make a new Residue object with the new atoms vector residues.addElement(new Residue(resAtoms, resNumber - 1, count)); Residue tmpres = residues.lastElement(); Atom tmpat = tmpres.atoms.get(0); // Make A new SequenceFeature for the current residue numbering SequenceFeature sf = new SequenceFeature("RESNUM", tmpat.resName + ":" + tmpat.resNumIns + " " + pdbid + id, "", offset + count, offset + count, pdbid); // MCview.PDBChain.PDBFILEFEATURE); resFeatures.addElement(sf); resAnnotation.addElement(new Annotation(tmpat.tfactor)); // Keep totting up the sequence if ((symbol = ResidueProperties.getAA3Hash().get(tmpat.resName)) == null) { String nucname = tmpat.resName.trim(); // use the aaIndex rather than call 'toLower' - which would take a bit // more time. deoxyn = nucname.length() == 2 && ResidueProperties.aaIndex[nucname.charAt(0)] == ResidueProperties.aaIndex['D']; if (tmpat.name.equalsIgnoreCase("CA") || ResidueProperties.nucleotideIndex[nucname .charAt((deoxyn ? 1 : 0))] == -1) { seq.append("X"); // System.err.println("PDBReader:Null aa3Hash for " + // tmpat.resName); } else { // nucleotide flag nucleotide = true; seq.append(nucname.charAt((deoxyn ? 1 : 0))); } } else { if (nucleotide) { System.err .println("Warning: mixed nucleotide and amino acid chain.. its gonna do bad things to you!"); } seq.append(ResidueProperties.aa[((Integer) symbol).intValue()]); } count++; } if (id.length() < 1) { id = " "; } isNa = nucleotide; sequence = new Sequence(id, seq.toString(), offset, resNumber - 1); // Note: // resNumber-offset // ~= // seq.size() // Add normalised feature scores to RESNUM indicating start/end of sequence // sf.setScore(offset+count); // System.out.println("PDB Sequence is :\nSequence = " + seq); // System.out.println("No of residues = " + residues.size()); for (i = 0, iSize = resFeatures.size(); i < iSize; i++) { sequence.addSequenceFeature(resFeatures.elementAt(i)); resFeatures.setElementAt(null, i); } if (visibleChainAnnotation) { Annotation[] annots = new Annotation[resAnnotation.size()]; float max = 0; for (i = 0, iSize = annots.length; i < iSize; i++) { annots[i] = resAnnotation.elementAt(i); if (annots[i].value > max) { max = annots[i].value; } resAnnotation.setElementAt(null, i); } AlignmentAnnotation tfactorann = new AlignmentAnnotation( "Temperature Factor", "Temperature Factor for " + pdbid + id, annots, 0, max, AlignmentAnnotation.LINE_GRAPH); tfactorann.setSequenceRef(sequence); sequence.addAlignmentAnnotation(tfactorann); } } /** * Colour start/end of bonds by charge *
    *
  • ASP and GLU red
  • *
  • LYS and ARG blue
  • *
  • CYS yellow
  • *
  • others light gray
  • *
*/ public void setChargeColours() { for (Bond b : bonds) { if (b.at1 != null && b.at2 != null) { b.startCol = getChargeColour(b.at1.resName); b.endCol = getChargeColour(b.at2.resName); } else { b.startCol = Color.gray; b.endCol = Color.gray; } } } public static Color getChargeColour(String resName) { Color result = Color.lightGray; if ("ASP".equals(resName) || "GLU".equals(resName)) { result = Color.red; } else if ("LYS".equals(resName) || "ARG".equals(resName)) { result = Color.blue; } else if ("CYS".equals(resName)) { result = Color.yellow; } return result; } /** * Sets the start/end colours of bonds to those of the start/end atoms * according to the specified colour scheme. Note: currently only works for * peptide residues. * * @param cs */ public void setChainColours(ColourSchemeI cs) { int index; for (Bond b : bonds) { try { index = ResidueProperties.aa3Hash.get(b.at1.resName) .intValue(); b.startCol = cs.findColour(ResidueProperties.aa[index].charAt(0)); index = ResidueProperties.aa3Hash.get(b.at2.resName) .intValue(); b.endCol = cs.findColour(ResidueProperties.aa[index].charAt(0)); } catch (Exception e) { b.startCol = Color.gray; b.endCol = Color.gray; } } } public void setChainColours(Color col) { for (Bond b : bonds) { b.startCol = col; b.endCol = col; } } /** * copy any sequence annotation onto the sequence mapped using the provided * StructureMapping * * @param mapping * - positional mapping between destination sequence and pdb resnum * @param sqmpping * - mapping between destination sequence and local chain */ public void transferResidueAnnotation(StructureMapping mapping, Mapping sqmpping) { SequenceI sq = mapping.getSequence(); SequenceI dsq = sq; if (sq != null) { while (dsq.getDatasetSequence() != null) { dsq = dsq.getDatasetSequence(); } // any annotation will be transferred onto the dataset sequence if (shadow != null && shadow.getAnnotation() != null) { for (AlignmentAnnotation ana : shadow.getAnnotation()) { List transfer = sq.getAlignmentAnnotations( ana.getCalcId(), ana.label); if (transfer == null || transfer.size() == 0) { ana = new AlignmentAnnotation(ana); ana.liftOver(sequence, shadowMap); ana.liftOver(dsq, sqmpping); dsq.addAlignmentAnnotation(ana); } else { continue; } } } else { if (sequence != null && sequence.getAnnotation() != null) { for (AlignmentAnnotation ana : sequence.getAnnotation()) { List transfer = sq .getAlignmentAnnotations(ana.getCalcId(), ana.label); if (transfer == null || transfer.size() == 0) { ana = new AlignmentAnnotation(ana); ana.liftOver(dsq, sqmpping); // mapping.transfer(ana); } else { continue; } } } } if (false) { // Useful for debugging mappings - adds annotation for mapped position float min = -1, max = 0; Annotation[] an = new Annotation[sq.getEnd() - sq.getStart() + 1]; for (int i = sq.getStart(), j = sq.getEnd(), k = 0; i <= j; i++, k++) { int prn = mapping.getPDBResNum(k + 1); an[k] = new Annotation(prn); if (min == -1) { min = k; max = k; } else { if (min > k) { min = k; } else if (max < k) { max = k; } } } sq.addAlignmentAnnotation(new AlignmentAnnotation("PDB.RESNUM", "PDB Residue Numbering for " + this.pdbid + ":" + this.id, an, min, max, AlignmentAnnotation.LINE_GRAPH)); } } } }