@PostConstruct
private void initDatabase() throws Exception {
if (ifCreateDropTables.equals("create-drop")) {
loadPeptideAnnotations();
loadReplicateAnnotations();
}
referenceP100Profile = repositoryService.getReferenceProfileVector(AssayType.P100);
referenceGCPProfile = repositoryService.getReferenceProfileVector(AssayType.GCP);
referenceP100GeneNames = repositoryService.getReferenceGeneNames(AssayType.P100);
referenceGCPGeneNames = repositoryService.getReferenceGeneNames(AssayType.GCP);
if (ifCreateDropTables.equals("create-drop")) {
// loadPeptideAnnotations();
// loadReplicateAnnotations();
loadDataPoints();
// normalize();
buildProfiles();
computeCorrelations();
mergeProfiles(referenceP100Profile.size(), referenceGCPProfile.size());
}
}
private void buildProfiles() {
log.info("Filling repository with profiles");
List<String> referenceProfile;
List<String> referenceGeneNames;
Set<GctReplicate> gctReplicatePairs = repositoryService.getGctReplicatesCombinations();
int dummyClusteringOrder = gctReplicatePairs.size();
for (GctReplicate gctReplicate : gctReplicatePairs) {
GctFile gctFile = gctReplicate.getGctFile();
ReplicateAnnotation replicateAnnotation = gctReplicate.getReplicateAnnotation();
List<PeakArea> peakAreas =
peakAreaRepository.findByGctFileAndReplicateAnnotation(gctFile, replicateAnnotation);
AssayType assayType = gctFile.getAssayType();
referenceProfile = getReferenceProfile(assayType);
referenceGeneNames = getReferenceGeneNames(assayType);
Double[] profileVector = new Double[referenceProfile.size()];
boolean[] imputeVector = new boolean[referenceProfile.size()];
for (PeakArea peakArea : peakAreas) {
int index = referenceProfile.indexOf(peakArea.getPeptideAnnotation().getPeptideId());
profileVector[index] = peakArea.getValue();
}
UtilsStatistics.imputeProfileVector(profileVector, imputeVector);
Profile profile =
new Profile(
replicateAnnotation,
gctFile,
ArrayUtils.toPrimitive(profileVector),
imputeVector,
referenceProfile,
referenceGeneNames,
dummyClusteringOrder--);
profileRepository.save(profile);
}
}
private void mergeProfiles(int p100Length, int gcpLength) {
log.info("Merging profiles");
ArrayList<AssayType> dummyAssay = new ArrayList<>();
dummyAssay.add(AssayType.GCP);
dummyAssay.add(AssayType.P100);
List<Profile> profiles = profileRepository.findByAssayTypeInOrderByConcatDesc(dummyAssay);
String prevConcat = null;
String curConcat;
List<Profile> bunchOfProfiles = null;
for (Profile profile : profiles) {
curConcat =
profile.getReplicateAnnotation().getCellId()
+ profile.getReplicateAnnotation().getPertiname();
if (prevConcat == null) {
prevConcat = curConcat;
bunchOfProfiles = new ArrayList<>();
}
if (!curConcat.equals(prevConcat)) {
prevConcat = curConcat;
MergedProfile mergedProfile =
UtilsTransform.mergeProfiles(bunchOfProfiles, p100Length, gcpLength);
mergedProfileRepository.save(mergedProfile);
bunchOfProfiles = new ArrayList<>();
} else {
bunchOfProfiles.add(profile);
}
}
}
private void computeCorrelations() {
log.info("Filling repository with most correlated profiles");
List<String> referenceProfile;
for (AssayType assayType : AssayType.values()) {
referenceProfile = getReferenceProfile(assayType);
List<Profile> profiles = profileRepository.findByAssayType(assayType);
String[] profileNames = new String[profiles.size()];
double[][] distanceMatrix = new double[profiles.size()][profiles.size()];
int i = 0;
for (Profile profileA : profiles) {
profileNames[i] = profileA.getId().toString();
Double maxPearson = Double.MIN_VALUE;
Profile maxProfile = profileA;
int j = 0;
for (Profile profileB : profiles) {
if (profileA.equals(profileB)) {
distanceMatrix[i][j] = 0;
j++;
continue;
}
double[] vectorA = profileA.getVector();
double[] vectorB = profileB.getVector();
PearsonsCorrelation pearson = new PearsonsCorrelation();
Double pearsonCorrelation = pearson.correlation(vectorA, vectorB);
if (pearsonCorrelation >= maxPearson) {
maxPearson = pearsonCorrelation;
maxProfile = profileB;
}
double[] profileAasDouble = UtilsTransform.intArrayToDouble(profileA.getColors());
double[] profileBasDouble = UtilsTransform.intArrayToDouble(profileB.getColors());
Double pearsonOfColors = pearson.correlation(profileAasDouble, profileBasDouble);
distanceMatrix[i][j] = pearsonOfColors;
j++;
}
profileA.setCorrelatedVector(maxProfile.getListWrapper());
SortedSet<StringDouble> positivePeptides =
UtilsStatistics.influentialPeptides(
profileA.getVector(), maxProfile.getVector(), referenceProfile, true);
profileA.setPositivePeptides(UtilsTransform.SortedSetToHTML(positivePeptides, false));
DecimalFormat df = new DecimalFormat("0.0000");
String peptideCorrelation = " <br/><br/><b style=\"color: #23527c;\">%s</b>";
profileA.setPositiveCorrelation(
maxProfile.toString() + String.format(peptideCorrelation, df.format(maxPearson)));
profileRepository.save(profileA);
i++;
}
}
}
private void normalize() {
List<AssayType> assayTypes = Arrays.asList(AssayType.GCP, AssayType.P100);
// do loop for each assay e.g. P100, GCP
for (AssayType assayType : assayTypes) {
log.info("Normalize matrix of peak values for assay: {}", assayType);
List<PeakArea> allPeakAreas = peakAreaRepository.findByGctFileAssayType(assayType);
// inefficient, try with database count
int numberOfPeptides = peptideAnnotationRepository.findByAssayType(assayType).size();
int numberOfReplicates = replicateAnnotationRepository.findByAssayType(assayType).size();
ArrayList<Integer> mapPeptideIdToRowId = new ArrayList<>();
ArrayList<Integer> mapReplicateIdToColumnId = new ArrayList<>();
// double[][] matrix = new double[replicates][peptides];
// init matrix with sizes
List<List<Double>> peaksAsMatrix = new ArrayList<>(numberOfReplicates);
for (int i = 0; i < numberOfReplicates; i++) {
peaksAsMatrix.add(new ArrayList<>(numberOfPeptides));
}
for (PeakArea peakArea : allPeakAreas) {
int peptideId = Math.toIntExact(peakArea.getPeptideAnnotation().getId());
int replicateId = Math.toIntExact(peakArea.getReplicateAnnotation().getId());
Double rawValue = peakArea.getValue();
int mappedRowId;
int mappedColumnId;
if (!mapPeptideIdToRowId.contains(peptideId)) {
mapPeptideIdToRowId.add(peptideId);
}
mappedRowId = mapPeptideIdToRowId.indexOf(peptideId);
if (!mapReplicateIdToColumnId.contains(replicateId)) {
mapReplicateIdToColumnId.add(replicateId);
}
mappedColumnId = mapReplicateIdToColumnId.indexOf(replicateId);
peaksAsMatrix.get(mappedColumnId).set(mappedRowId, rawValue);
}
List<List<Double>> outputMatrix = Normalizer.quantileAndZScoreNormalize(peaksAsMatrix);
// write normalized values back to DB
for (int i = 0; i < numberOfReplicates; i++) {
for (int j = 0; j < numberOfPeptides; j++) {
Double normalizedValue = outputMatrix.get(i).get(j);
Long databaseReplicateId = mapReplicateIdToColumnId.get(i).longValue();
Long databasePeptideId = mapPeptideIdToRowId.get(j).longValue();
List<PeakArea> peakAreas =
peakAreaRepository
.findByGctFileAssayTypeAndReplicateAnnotationIdAndPeptideAnnotationId(
assayType, databaseReplicateId, databasePeptideId);
assert peakAreas.size() == 1;
PeakArea peakArea = peakAreas.get(0);
peakArea.setNormalizedValue(normalizedValue);
// Fill in normalized value field in DB
peakAreaRepository.save(peakArea);
}
}
log.info(
"Normalized assay: {} peptides: {}, replicates: {}.",
assayType,
numberOfPeptides,
numberOfReplicates);
}
}