/**
* Takes a potentially non-sequential alignment and guesses a sequential version of it. Residues
* from each structure are sorted sequentially and then compared directly.
*
* <p>The results of this method are consistent with what one might expect from an identity
* function, and are therefore useful with {@link #getSymmetryOrder(Map, Map identity, int,
* float)}.
*
* <ul>
* <li>Perfect self-alignments will have the same pre-image and image, so will map X->X
* <li>Gaps and alignment errors will cause errors in the resulting map, but only locally.
* Errors do not propagate through the whole alignment.
* </ul>
*
* <h4>Example:</h4>
*
* A non sequential alignment, represented schematically as
*
* <pre>
* 12456789
* 78912345</pre>
*
* would result in a map
*
* <pre>
* 12456789
* 12345789</pre>
*
* @param alignment The non-sequential input alignment
* @param inverseAlignment If false, map from structure1 to structure2. If true, generate the
* inverse of that map.
* @return A mapping from sequential residues of one protein to those of the other
* @throws IllegalArgumentException if the input alignment is not one-to-one.
*/
public static Map<Integer, Integer> guessSequentialAlignment(
Map<Integer, Integer> alignment, boolean inverseAlignment) {
Map<Integer, Integer> identity = new HashMap<Integer, Integer>();
SortedSet<Integer> aligned1 = new TreeSet<Integer>();
SortedSet<Integer> aligned2 = new TreeSet<Integer>();
for (Entry<Integer, Integer> pair : alignment.entrySet()) {
aligned1.add(pair.getKey());
if (!aligned2.add(pair.getValue()))
throw new IllegalArgumentException(
"Alignment is not one-to-one for residue "
+ pair.getValue()
+ " of the second structure.");
}
Iterator<Integer> it1 = aligned1.iterator();
Iterator<Integer> it2 = aligned2.iterator();
while (it1.hasNext()) {
if (inverseAlignment) { // 2->1
identity.put(it2.next(), it1.next());
} else { // 1->2
identity.put(it1.next(), it2.next());
}
}
return identity;
}
/** @see #toConciseAlignmentString(Map, Map) */
public static Map<Integer, Integer> fromConciseAlignmentString(String string) {
Map<Integer, Integer> map = new HashMap<Integer, Integer>();
boolean matches = true;
while (matches) {
Pattern pattern = Pattern.compile("(\\d+)>(\\d+)");
Matcher matcher = pattern.matcher(string);
matches = matcher.find();
if (matches) {
Integer from = Integer.parseInt(matcher.group(1));
Integer to = Integer.parseInt(matcher.group(2));
map.put(from, to);
string = string.substring(matcher.end(1) + 1);
}
}
return map;
}
/**
* Applies an alignment k times. Eg if alignmentMap defines function f(x), this returns a function
* f^k(x)=f(f(...f(x)...)).
*
* <p>To allow for functions with different domains and codomains, the identity function allows
* converting back in a reasonable way. For instance, if alignmentMap represented an alignment
* between two proteins with different numbering schemes, the identity function could calculate
* the offset between residue numbers, eg I(x) = x-offset.
*
* <p>When an identity function is provided, the returned function calculates f^k(x) = f(I( f(I(
* ... f(x) ... )) )).
*
* @param <S>
* @param <T>
* @param alignmentMap The input function, as a map (see {@link
* AlignmentTools#alignmentAsMap(AFPChain)})
* @param identity An identity-like function providing the isomorphism between the codomain of
* alignmentMap (of type <T>) and the domain (type <S>).
* @param k The number of times to apply the alignment
* @return A new alignment. If the input function is not automorphic (one-to-one), then some
* inputs may map to null, indicating that the function is undefined for that input.
*/
public static <S, T> Map<S, T> applyAlignment(Map<S, T> alignmentMap, Map<T, S> identity, int k) {
// This implementation simply applies the map k times.
// If k were large, it would be more efficient to do this recursively,
// (eg f^4 = (f^2)^2) but k will usually be small.
if (k < 0) throw new IllegalArgumentException("k must be positive");
if (k == 1) {
return new HashMap<S, T>(alignmentMap);
}
// Convert to lists to establish a fixed order
List<S> preimage = new ArrayList<S>(alignmentMap.keySet()); // currently unmodified
List<S> image = new ArrayList<S>(preimage);
for (int n = 1; n < k; n++) {
// apply alignment
for (int i = 0; i < image.size(); i++) {
S pre = image.get(i);
T intermediate = (pre == null ? null : alignmentMap.get(pre));
S post = (intermediate == null ? null : identity.get(intermediate));
image.set(i, post);
}
}
Map<S, T> imageMap = new HashMap<S, T>(alignmentMap.size());
// TODO handle nulls consistently.
// assure that all the residues in the domain are valid keys
/*
for(int i=0;i<preimage.size();i++) {
S pre = preimage.get(i);
T intermediate = (pre==null?null: alignmentMap.get(pre));
S post = (intermediate==null?null: identity.get(intermediate));
imageMap.put(post, null);
}
*/
// now populate with actual values
for (int i = 0; i < preimage.size(); i++) {
S pre = preimage.get(i);
// image is currently f^k-1(x), so take the final step
S preK1 = image.get(i);
T postK = (preK1 == null ? null : alignmentMap.get(preK1));
imageMap.put(pre, postK);
}
return imageMap;
}
/**
* Creates a Map specifying the alignment as a mapping between residue indices of protein 1 and
* residue indices of protein 2.
*
* <p>For example,
*
* <pre>
* 1234
* 5678</pre>
*
* becomes
*
* <pre>
* 1->5
* 2->6
* 3->7
* 4->8</pre>
*
* @param afpChain An alignment
* @return A mapping from aligned residues of protein 1 to their partners in protein 2.
* @throws StructureException If afpChain is not one-to-one
*/
public static Map<Integer, Integer> alignmentAsMap(AFPChain afpChain) throws StructureException {
Map<Integer, Integer> map = new HashMap<Integer, Integer>();
if (afpChain.getAlnLength() < 1) {
return map;
}
int[][][] optAln = afpChain.getOptAln();
int[] optLen = afpChain.getOptLen();
for (int block = 0; block < afpChain.getBlockNum(); block++) {
for (int pos = 0; pos < optLen[block]; pos++) {
int res1 = optAln[block][0][pos];
int res2 = optAln[block][1][pos];
if (map.containsKey(res1)) {
throw new StructureException(
String.format("Residue %d aligned to both %d and %d.", res1, map.get(res1), res2));
}
map.put(res1, res2);
}
}
return map;
}
/**
* Identify a set of modifications in a a list of chains.
*
* @param chains query {@link Chain}s.
* @param potentialModifications query {@link ProteinModification}s.
*/
public void identify(
final List<Chain> chains, final Set<ProteinModification> potentialModifications) {
if (chains == null) {
throw new IllegalArgumentException("Null structure.");
}
if (potentialModifications == null) {
throw new IllegalArgumentException("Null potentialModifications.");
}
reset();
if (potentialModifications.isEmpty()) {
return;
}
Map<String, Chain> mapChainIdChain = new HashMap<String, Chain>(chains.size());
residues = new ArrayList<Group>();
List<Group> ligands = new ArrayList<Group>();
Map<Component, Set<Group>> mapCompGroups = new HashMap<Component, Set<Group>>();
for (Chain chain : chains) {
mapChainIdChain.put(chain.getChainID(), chain);
List<Group> ress = StructureUtil.getAminoAcids(chain);
// List<Group> ligs = chain.getAtomLigands();
List<Group> ligs = StructureTools.filterLigands(chain.getAtomGroups());
residues.addAll(ress);
residues.removeAll(ligs);
ligands.addAll(ligs);
addModificationGroups(potentialModifications, ress, ligs, mapCompGroups);
}
if (residues.isEmpty()) {
String pdbId = "?";
if (chains.size() > 0) {
Structure struc = chains.get(0).getParent();
if (struc != null) pdbId = struc.getPDBCode();
}
logger.warn(
"No amino acids found for {}. Either you did not parse the PDB file with alignSEQRES records, or this record does not contain any amino acids.",
pdbId);
}
List<ModifiedCompound> modComps = new ArrayList<ModifiedCompound>();
for (ProteinModification mod : potentialModifications) {
ModificationCondition condition = mod.getCondition();
List<Component> components = condition.getComponents();
if (!mapCompGroups.keySet().containsAll(components)) {
// not all components exist for this mod.
continue;
}
int sizeComps = components.size();
if (sizeComps == 1) {
processCrosslink1(mapCompGroups, modComps, mod, components);
} else {
processMultiCrosslink(mapCompGroups, modComps, mod, condition);
}
}
if (recordAdditionalAttachments) {
// identify additional groups that are not directly attached to amino acids.
for (ModifiedCompound mc : modComps) {
identifyAdditionalAttachments(mc, ligands, mapChainIdChain);
}
}
mergeModComps(modComps);
identifiedModifiedCompounds.addAll(modComps);
// record unidentifiable linkage
if (recordUnidentifiableModifiedCompounds) {
recordUnidentifiableAtomLinkages(modComps, ligands);
recordUnidentifiableModifiedResidues(modComps);
}
}