/**
* Print an alignment map in a concise representation. Edges are given as two numbers separated by
* '>'. They are chained together where possible, or separated by spaces where disjoint or
* branched.
*
* <p>Note that more concise representations may be possible. Examples:
* <li>1>2>3>1
* <li>1>2>3>2 4>3
*
* @param alignment The input function, as a map (see {@link
* AlignmentTools#alignmentAsMap(AFPChain)})
* @param identity An identity-like function providing the isomorphism between the codomain of
* alignment (of type <T>) and the domain (type <S>).
* @return
*/
public static <S, T> String toConciseAlignmentString(Map<S, T> alignment, Map<T, S> identity) {
// Clone input to prevent changes
Map<S, T> alig = new HashMap<S, T>(alignment);
// Generate inverse alignment
Map<S, List<S>> inverse = new HashMap<S, List<S>>();
for (Entry<S, T> e : alig.entrySet()) {
S val = identity.get(e.getValue());
if (inverse.containsKey(val)) {
List<S> l = inverse.get(val);
l.add(e.getKey());
} else {
List<S> l = new ArrayList<S>();
l.add(e.getKey());
inverse.put(val, l);
}
}
StringBuilder str = new StringBuilder();
while (!alig.isEmpty()) {
// Pick an edge and work upstream to a root or cycle
S seedNode = alig.keySet().iterator().next();
S node = seedNode;
if (inverse.containsKey(seedNode)) {
node = inverse.get(seedNode).iterator().next();
while (node != seedNode && inverse.containsKey(node)) {
node = inverse.get(node).iterator().next();
}
}
// Now work downstream, deleting edges as we go
seedNode = node;
str.append(node);
while (alig.containsKey(node)) {
S lastNode = node;
node = identity.get(alig.get(lastNode));
// Output
str.append('>');
str.append(node);
// Remove edge
alig.remove(lastNode);
List<S> inv = inverse.get(node);
if (inv.size() > 1) {
inv.remove(node);
} else {
inverse.remove(node);
}
}
if (!alig.isEmpty()) {
str.append(' ');
}
}
return str.toString();
}
/**
* Builds an {@link AFPChain} from already-matched arrays of atoms and residues.
*
* @param ca1 An array of atoms in the first structure
* @param ca2 An array of atoms in the second structure
* @param residues1 An array of {@link ResidueNumber ResidueNumbers} in the first structure that
* are aligned. Only null ResidueNumbers are considered to be unaligned
* @param residues2 An array of {@link ResidueNumber ResidueNumbers} in the second structure that
* are aligned. Only null ResidueNumbers are considered to be unaligned
* @throws StructureException
*/
private static AFPChain buildAlignment(
Atom[] ca1, Atom[] ca2, ResidueNumber[] residues1, ResidueNumber[] residues2)
throws StructureException {
// remove any gap
// this includes the ones introduced by the nullifying above
List<ResidueNumber> alignedResiduesList1 = new ArrayList<ResidueNumber>();
List<ResidueNumber> alignedResiduesList2 = new ArrayList<ResidueNumber>();
for (int i = 0; i < residues1.length; i++) {
if (residues1[i] != null && residues2[i] != null) {
alignedResiduesList1.add(residues1[i]);
alignedResiduesList2.add(residues2[i]);
}
}
ResidueNumber[] alignedResidues1 =
alignedResiduesList1.toArray(new ResidueNumber[alignedResiduesList1.size()]);
ResidueNumber[] alignedResidues2 =
alignedResiduesList2.toArray(new ResidueNumber[alignedResiduesList2.size()]);
AFPChain afpChain = AlignmentTools.createAFPChain(ca1, ca2, alignedResidues1, alignedResidues2);
afpChain.setAlgorithmName("unknown");
AlignmentTools.updateSuperposition(afpChain, ca1, ca2);
afpChain.setBlockSize(new int[] {afpChain.getNrEQR()});
afpChain.setBlockRmsd(new double[] {afpChain.getTotalRmsdOpt()});
afpChain.setBlockGap(new int[] {afpChain.getGapLen()});
return afpChain;
}
/**
* identify additional groups that are not directly attached to amino acids.
*
* @param mc {@link ModifiedCompound}.
* @param chain a {@link Chain}.
* @return a list of added groups.
*/
private void identifyAdditionalAttachments(
ModifiedCompound mc, List<Group> ligands, Map<String, Chain> mapChainIdChain) {
if (ligands.isEmpty()) {
return;
}
// TODO: should the additional groups only be allowed to the identified
// ligands or both amino acids and ligands? Currently only on ligands
// ligands to amino acid bonds for same modification of unknown category
// will be combined in mergeModComps()
// TODO: how about chain-chain links?
List<Group> identifiedGroups = new ArrayList<Group>();
for (StructureGroup num : mc.getGroups(false)) {
Group group;
try {
// String numIns = "" + num.getResidueNumber();
// if (num.getInsCode() != null) {
// numIns += num.getInsCode();
// }
ResidueNumber resNum = new ResidueNumber();
resNum.setChainId(num.getChainId());
resNum.setSeqNum(num.getResidueNumber());
resNum.setInsCode(num.getInsCode());
// group = chain.getGroupByPDB(numIns);
group = mapChainIdChain.get(num.getChainId()).getGroupByPDB(resNum);
} catch (StructureException e) {
logger.error("Exception: ", e);
// should not happen
continue;
}
identifiedGroups.add(group);
}
int start = 0;
int n = identifiedGroups.size();
while (n > start) {
for (Group group1 : ligands) {
for (int i = start; i < n; i++) {
Group group2 = identifiedGroups.get(i);
if (!identifiedGroups.contains(group1)) {
List<Atom[]> linkedAtoms =
StructureUtil.findAtomLinkages(group1, group2, false, bondLengthTolerance);
if (!linkedAtoms.isEmpty()) {
for (Atom[] atoms : linkedAtoms) {
mc.addAtomLinkage(
StructureUtil.getStructureAtomLinkage(atoms[0], false, atoms[1], false));
}
identifiedGroups.add(group1);
break;
}
}
}
}
start = n;
n = identifiedGroups.size();
}
}
/**
* Takes an AFPChain and replaces the optimal alignment based on an alignment map
*
* <p>Parameters are filled with defaults (often null) or sometimes calculated.
*
* <p>For a way to create a new AFPChain, see {@link AlignmentTools#createAFPChain(Atom[], Atom[],
* ResidueNumber[], ResidueNumber[])}
*
* @param afpChain The alignment to be modified
* @param alignment The new alignment, as a Map
* @throws StructureException if an error occurred during superposition
* @see AlignmentTools#createAFPChain(Atom[], Atom[], ResidueNumber[], ResidueNumber[])
*/
public static AFPChain replaceOptAln(
AFPChain afpChain, Atom[] ca1, Atom[] ca2, Map<Integer, Integer> alignment)
throws StructureException {
// Determine block lengths
// Sort ca1 indices, then start a new block whenever ca2 indices aren't
// increasing monotonically.
Integer[] res1 = alignment.keySet().toArray(new Integer[0]);
Arrays.sort(res1);
List<Integer> blockLens = new ArrayList<Integer>(2);
int optLength = 0;
Integer lastRes = alignment.get(res1[0]);
int blkLen = lastRes == null ? 0 : 1;
for (int i = 1; i < res1.length; i++) {
Integer currRes = alignment.get(res1[i]); // res2 index
assert (currRes
!= null); // could be converted to if statement if assertion doesn't hold; just modify
// below as well.
if (lastRes < currRes) {
blkLen++;
} else {
// CP!
blockLens.add(blkLen);
optLength += blkLen;
blkLen = 1;
}
lastRes = currRes;
}
blockLens.add(blkLen);
optLength += blkLen;
// Create array structure for alignment
int[][][] optAln = new int[blockLens.size()][][];
int pos1 = 0; // index into res1
for (int blk = 0; blk < blockLens.size(); blk++) {
optAln[blk] = new int[2][];
blkLen = blockLens.get(blk);
optAln[blk][0] = new int[blkLen];
optAln[blk][1] = new int[blkLen];
int pos = 0; // index into optAln
while (pos < blkLen) {
optAln[blk][0][pos] = res1[pos1];
Integer currRes = alignment.get(res1[pos1]);
optAln[blk][1][pos] = currRes;
pos++;
pos1++;
}
}
assert (pos1 == optLength);
// Create length array
int[] optLens = new int[blockLens.size()];
for (int i = 0; i < blockLens.size(); i++) {
optLens[i] = blockLens.get(i);
}
return replaceOptAln(afpChain, ca1, ca2, blockLens.size(), optLens, optAln);
}
/**
* Assembly the matched linkages.
*
* @param matchedAtomsOfLinkages
* @param mod
* @param condition
* @param ret ModifiedCompound will be stored here.
*/
private void assembleLinkages(
List<List<Atom[]>> matchedAtomsOfLinkages,
ProteinModification mod,
List<ModifiedCompound> ret) {
ModificationCondition condition = mod.getCondition();
List<ModificationLinkage> modLinks = condition.getLinkages();
int nLink = matchedAtomsOfLinkages.size();
int[] indices = new int[nLink];
Set<ModifiedCompound> identifiedCompounds = new HashSet<ModifiedCompound>();
while (indices[0] < matchedAtomsOfLinkages.get(0).size()) {
List<Atom[]> atomLinkages = new ArrayList<Atom[]>(nLink);
for (int iLink = 0; iLink < nLink; iLink++) {
Atom[] atoms = matchedAtomsOfLinkages.get(iLink).get(indices[iLink]);
atomLinkages.add(atoms);
}
if (matchLinkages(modLinks, atomLinkages)) {
// matched
int n = atomLinkages.size();
List<StructureAtomLinkage> linkages = new ArrayList<StructureAtomLinkage>(n);
for (int i = 0; i < n; i++) {
Atom[] linkage = atomLinkages.get(i);
StructureAtomLinkage link =
StructureUtil.getStructureAtomLinkage(
linkage[0], residues.contains(linkage[0].getGroup()),
linkage[1], residues.contains(linkage[1].getGroup()));
linkages.add(link);
}
ModifiedCompound mc = new ModifiedCompoundImpl(mod, linkages);
if (!identifiedCompounds.contains(mc)) {
ret.add(mc);
identifiedCompounds.add(mc);
}
}
// indices++ (e.g. [0,0,1]=>[0,0,2]=>[1,2,0])
int i = nLink - 1;
while (i >= 0) {
if (i == 0 || indices[i] < matchedAtomsOfLinkages.get(i).size() - 1) {
indices[i]++;
break;
} else {
indices[i] = 0;
i--;
}
}
}
}
/**
* Takes a protein sequence string with capital and lowercase letters and sets its {@link
* ProteinSequence#getUserCollection() user collection} to record which letters are uppercase
* (aligned) and which are lowercase (unaligned).
*
* @param sequence Make sure <em>not</em> to use {@link ProteinSequence#getSequenceAsString()} for
* this, as it won't preserve upper- and lower-case
*/
public static List<Object> getAlignedUserCollection(String sequence) {
List<Object> aligned = new ArrayList<Object>(sequence.length());
for (char c : sequence.toCharArray()) {
aligned.add(Character.isUpperCase(c));
}
return aligned;
}
/**
* Retrieves the optimum alignment from an AFPChain and returns it as a java collection. The
* result is indexed in the same way as {@link AFPChain#getOptAln()}, but has the correct size().
*
* <pre>
* List<List<List<Integer>>> aln = getOptAlnAsList(AFPChain afpChain);
* aln.get(blockNum).get(structureNum={0,1}).get(pos)</pre>
*
* @param afpChain
* @return
*/
public static List<List<List<Integer>>> getOptAlnAsList(AFPChain afpChain) {
int[][][] optAln = afpChain.getOptAln();
int[] optLen = afpChain.getOptLen();
List<List<List<Integer>>> blocks = new ArrayList<List<List<Integer>>>(afpChain.getBlockNum());
for (int blockNum = 0; blockNum < afpChain.getBlockNum(); blockNum++) {
// TODO could improve speed an memory by wrapping the arrays with
// an unmodifiable list, similar to Arrays.asList(...) but with the
// correct size parameter.
List<Integer> align1 = new ArrayList<Integer>(optLen[blockNum]);
List<Integer> align2 = new ArrayList<Integer>(optLen[blockNum]);
for (int pos = 0; pos < optLen[blockNum]; pos++) {
align1.add(optAln[blockNum][0][pos]);
align2.add(optAln[blockNum][1][pos]);
}
List<List<Integer>> block = new ArrayList<List<Integer>>(2);
block.add(align1);
block.add(align2);
blocks.add(block);
}
return blocks;
}
/**
* Uses two sequences each with a corresponding structure to create an AFPChain corresponding to
* the alignment. Provided only for convenience since FastaReaders return such maps.
*
* @param sequences A Map containing exactly two entries from sequence names as Strings to gapped
* ProteinSequences; the name is ignored
* @see #fastaToAfpChain(ProteinSequence, ProteinSequence, Structure, Structure)
* @throws StructureException
*/
public static AFPChain fastaToAfpChain(
Map<String, ProteinSequence> sequences, Structure structure1, Structure structure2)
throws StructureException {
if (sequences.size() != 2) {
throw new IllegalArgumentException(
"There must be exactly 2 sequences, but there were " + sequences.size());
}
if (structure1 == null || structure2 == null) {
throw new IllegalArgumentException("A structure is null");
}
List<ProteinSequence> seqs = new ArrayList<ProteinSequence>();
List<String> names = new ArrayList<String>(2);
for (Map.Entry<String, ProteinSequence> entry : sequences.entrySet()) {
seqs.add(entry.getValue());
names.add(entry.getKey());
}
return fastaToAfpChain(seqs.get(0), seqs.get(1), structure1, structure2);
}
private void processCrosslink1(
Map<Component, Set<Group>> mapCompGroups,
List<ModifiedCompound> modComps,
ProteinModification mod,
List<Component> components) {
// modified residue
// TODO: is this the correct logic for CROSS_LINK_1?
Set<Group> modifiedResidues = mapCompGroups.get(components.get(0));
if (modifiedResidues != null) {
for (Group residue : modifiedResidues) {
StructureGroup strucGroup = StructureUtil.getStructureGroup(residue, true);
ModifiedCompound modRes = new ModifiedCompoundImpl(mod, strucGroup);
modComps.add(modRes);
}
}
}
/**
* @param a
* @param ca1
* @param ca2
* @return
* @throws StructureException if an error occurred during superposition
*/
public static AFPChain splitBlocksByTopology(AFPChain a, Atom[] ca1, Atom[] ca2)
throws StructureException {
int[][][] optAln = a.getOptAln();
int blockNum = a.getBlockNum();
int[] optLen = a.getOptLen();
// Determine block lengths
// Split blocks if residue indices don't increase monotonically
List<Integer> newBlkLen = new ArrayList<Integer>();
boolean blockChanged = false;
for (int blk = 0; blk < blockNum; blk++) {
int currLen = 1;
for (int pos = 1; pos < optLen[blk]; pos++) {
if (optAln[blk][0][pos] <= optAln[blk][0][pos - 1]
|| optAln[blk][1][pos] <= optAln[blk][1][pos - 1]) {
// start a new block
newBlkLen.add(currLen);
currLen = 0;
blockChanged = true;
}
currLen++;
}
if (optLen[blk] < 2) {
newBlkLen.add(optLen[blk]);
} else {
newBlkLen.add(currLen);
}
}
// Check if anything needs to be split
if (!blockChanged) {
return a;
}
// Split blocks
List<int[][]> blocks = new ArrayList<int[][]>(newBlkLen.size());
int oldBlk = 0;
int pos = 0;
for (int blkLen : newBlkLen) {
if (blkLen == optLen[oldBlk]) {
assert (pos == 0); // should be the whole block
// Use the old block
blocks.add(optAln[oldBlk]);
} else {
int[][] newBlock = new int[2][blkLen];
assert (pos + blkLen <= optLen[oldBlk]); // don't overrun block
for (int i = 0; i < blkLen; i++) {
newBlock[0][i] = optAln[oldBlk][0][pos + i];
newBlock[1][i] = optAln[oldBlk][1][pos + i];
}
pos += blkLen;
blocks.add(newBlock);
if (pos == optLen[oldBlk]) {
// Finished this oldBlk, start the next
oldBlk++;
pos = 0;
}
}
}
// Store new blocks
int[][][] newOptAln = blocks.toArray(new int[0][][]);
int[] newBlockLens = new int[newBlkLen.size()];
for (int i = 0; i < newBlkLen.size(); i++) {
newBlockLens[i] = newBlkLen.get(i);
}
return replaceOptAln(a, ca1, ca2, blocks.size(), newBlockLens, newOptAln);
}
/** Get matched atoms for all linkages. */
private List<List<Atom[]>> getMatchedAtomsOfLinkages(
ModificationCondition condition, Map<Component, Set<Group>> mapCompGroups) {
List<ModificationLinkage> linkages = condition.getLinkages();
int nLink = linkages.size();
List<List<Atom[]>> matchedAtomsOfLinkages = new ArrayList<List<Atom[]>>(nLink);
for (int iLink = 0; iLink < nLink; iLink++) {
ModificationLinkage linkage = linkages.get(iLink);
Component comp1 = linkage.getComponent1();
Component comp2 = linkage.getComponent2();
// boolean isAA1 = comp1.;
// boolean isAA2 = comp2.getType()==true;
Set<Group> groups1 = mapCompGroups.get(comp1);
Set<Group> groups2 = mapCompGroups.get(comp2);
List<Atom[]> list = new ArrayList<Atom[]>();
List<String> potentialNamesOfAtomOnGroup1 = linkage.getPDBNameOfPotentialAtomsOnComponent1();
for (String name : potentialNamesOfAtomOnGroup1) {
if (name.equals("*")) {
// wildcard
potentialNamesOfAtomOnGroup1 = null; // search all atoms
break;
}
}
List<String> potentialNamesOfAtomOnGroup2 = linkage.getPDBNameOfPotentialAtomsOnComponent2();
for (String name : potentialNamesOfAtomOnGroup2) {
if (name.equals("*")) {
// wildcard
potentialNamesOfAtomOnGroup2 = null; // search all atoms
break;
}
}
for (Group g1 : groups1) {
for (Group g2 : groups2) {
if (g1.equals(g2)) {
continue;
}
// only for wildcard match of two residues
boolean ignoreNCLinkage =
potentialNamesOfAtomOnGroup1 == null
&& potentialNamesOfAtomOnGroup2 == null
&& residues.contains(g1)
&& residues.contains(g2);
Atom[] atoms =
StructureUtil.findNearestAtomLinkage(
g1,
g2,
potentialNamesOfAtomOnGroup1,
potentialNamesOfAtomOnGroup2,
ignoreNCLinkage,
bondLengthTolerance);
if (atoms != null) {
list.add(atoms);
}
}
}
if (list.isEmpty()) {
// broken linkage
break;
}
matchedAtomsOfLinkages.add(list);
}
return matchedAtomsOfLinkages;
}