/**
* @param orConditions StatisticsQueryOrConditions<StatisticsQueryCondition>
* @param statisticsStorage
* @param scoringExps Set of experiments that have at least one non-zero score for
* statisticsQuery. This is used retrieving efos to be displayed in heatmap when no query efvs
* exist (c.f. atlasStatisticsQueryService.getScoringAttributesForGenes())
* @return Multiset<Integer> containing experiment counts corresponding to all attributes in each
* StatisticsQueryCondition in orConditions
*/
private static Multiset<Integer> getScoresForOrConditions(
final StatisticsQueryOrConditions<StatisticsQueryCondition> orConditions,
StatisticsStorage statisticsStorage,
Set<ExperimentInfo> scoringExps) {
Multiset<Integer> scores = HashMultiset.create();
for (StatisticsQueryCondition orCondition : orConditions.getConditions()) {
orCondition.setBioEntityIdRestrictionSet(orConditions.getBioEntityIdRestrictionSet());
scores.addAll(scoreQuery(orCondition, statisticsStorage, scoringExps));
}
// Now apply orConditions' min experiments restriction to scores
Multiset<Integer> qualifyingScores = HashMultiset.create();
for (Multiset.Entry<Integer> entry : scores.entrySet()) {
if (entry.getCount() >= orConditions.getMinExperiments()) {
qualifyingScores.setCount(entry.getElement(), entry.getCount());
}
}
return qualifyingScores;
}
/**
* @param orAttributes
* @param statType
* @param minExperiments minimum experiment count restriction for this clause
* @param statisticsStorage - used to retrieve orAttributes, needed finding experiment counts in
* bit index
* @return StatisticsQueryOrConditions representing orAttributes
*/
public static StatisticsQueryOrConditions<StatisticsQueryCondition> getStatisticsOrQuery(
final List<Attribute> orAttributes,
final StatisticsType statType,
int minExperiments,
final StatisticsStorage statisticsStorage) {
StatisticsQueryOrConditions<StatisticsQueryCondition> orConditions =
new StatisticsQueryOrConditions<StatisticsQueryCondition>();
orConditions.setMinExperiments(minExperiments);
// LinkedHashMap used to maintain ordering of processing of experiments in multi-Attribute,
// multi-Experiment bit index queries to
// retrieve sorted lists of experiments to be plotted on the gene page.
Map<ExperimentInfo, Set<EfAttribute>> allExpsToAttrs =
new LinkedHashMap<ExperimentInfo, Set<EfAttribute>>();
for (Attribute attr : orAttributes) {
attr.getAttributeToExperimentMappings(statisticsStorage, allExpsToAttrs);
}
// Now process allExpsToAttrs - for all efo terms in orAttributes, grouping into one
// StatisticsQueryCondition
// attributes from potentially different efoTerms for one experiment. This has the effect of
// counting a given
// experiment only once for an OR collection of Attributes.
for (Map.Entry<ExperimentInfo, Set<EfAttribute>> expToAttr : allExpsToAttrs.entrySet()) {
StatisticsQueryCondition cond = new StatisticsQueryCondition(statType);
if (expToAttr.getKey() != EfAttribute.ALL_EXPERIMENTS)
// For efv Attributes we span all experiments
cond.inExperiments(Collections.singletonList(expToAttr.getKey()));
for (EfAttribute attr : expToAttr.getValue()) {
cond.inAttribute(attr);
}
orConditions.orCondition(cond);
}
return orConditions;
}
/**
* The core scoring method for statistics queries
*
* @param statisticsQuery query to be peformed on statisticsStorage
* @param statisticsStorage core data for Statistics qeries
* @param scoringExps an out parameter.
* <p>- If null, experiment counts result of statisticsQuery should be returned. if - If
* non-null, it serves as a flag that an optimised statisticsQuery should be performed to just
* collect Experiments for which non-zero counts exist for Statistics query. A typical call
* scenario in this case is just one efv per statisticsQuery, in which we can both: 1. check
* if the efv Attribute itself is a scoring one 2. map this Attribute and Experimeants in
* scoringExps to efo terms - via the reverse mapping efv-experiment-> efo term in EfoIndex
* (c.f. atlasStatisticsQueryService.getScoringAttributesForGenes())
* @return Multiset of aggregated experiment counts, where the set of scores genes is intersected
* across statisticsQuery.getConditions(), and union-ed across attributes within each
* condition in statisticsQuery.getConditions().
*/
public static Multiset<Integer> scoreQuery(
StatisticsQueryCondition statisticsQuery,
final StatisticsStorage statisticsStorage,
Set<ExperimentInfo> scoringExps) {
// gatherScoringExpsOnly -> experiment counts should be calculated for statisticsQuery
// !gatherScoringExpsOnly -> scoring experiments should be collected (into scoringExps) only
boolean gatherScoringExpsOnly = scoringExps != null;
Set<StatisticsQueryOrConditions<StatisticsQueryCondition>> andStatisticsQueryConditions =
statisticsQuery.getConditions();
Multiset<Integer> results = null;
if (andStatisticsQueryConditions.isEmpty()) { // End of recursion
Set<Integer> bioEntityIdRestrictionSet = statisticsQuery.getBioEntityIdRestrictionSet();
Set<EfAttribute> attributes = statisticsQuery.getAttributes();
if (attributes.isEmpty()) {
// No attributes were provided - we have to use pre-computed scores across all attributes
Multiset<Integer> scoresAcrossAllEfos =
statisticsStorage.getScoresAcrossAllEfos(statisticsQuery.getStatisticsType());
results = intersect(scoresAcrossAllEfos, bioEntityIdRestrictionSet);
} else {
results = HashMultiset.create();
setQueryExperiments(statisticsQuery, statisticsStorage);
// For each experiment in the query, traverse through all attributes and add all gene
// indexes into one ConciseSet. This way a gene can score
// only once for a single experiment - across all OR attributes in this query. Once all
// attributes have been traversed for a single experiment,
// add ConciseSet to Multiset results
for (ExperimentInfo exp : statisticsQuery.getExperiments()) {
FastSet statsForExperiment = new FastSet();
for (EfAttribute attr : attributes) {
Map<ExperimentInfo, ConciseSet> expsToStats =
getStatisticsForAttribute(
statisticsQuery.getStatisticsType(), attr, statisticsStorage);
if (expsToStats != null) {
if (expsToStats.isEmpty()) {
log.debug(
"Failed to retrieve stats for stat: "
+ statisticsQuery.getStatisticsType()
+ " and attr: "
+ attr);
} else {
if (expsToStats.get(exp) != null) {
if (!gatherScoringExpsOnly) {
statsForExperiment.addAll(
intersect(expsToStats.get(exp), bioEntityIdRestrictionSet));
} else if (containsAtLeastOne(expsToStats.get(exp), bioEntityIdRestrictionSet)) {
// exp contains at least one non-zero score for at least one gene index in
// bioEntityIdRestrictionSet -> add it to scoringExps
scoringExps.add(exp);
}
} else {
log.debug(
"Failed to retrieve stats for stat: "
+ statisticsQuery.getStatisticsType()
+ " exp: "
+ exp.getAccession()
+ " and attr: "
+ attr);
}
}
}
}
if (!gatherScoringExpsOnly) {
results.addAll(statsForExperiment);
}
}
}
} else {
// run over all AND conditions, do "OR" inside (cf. scoreOrStatisticsQueryConditions()) ,
// "AND"'ing over the whole thing
for (StatisticsQueryOrConditions<StatisticsQueryCondition> orConditions :
andStatisticsQueryConditions) {
// Pass gene restriction set down to orConditions
orConditions.setGeneRestrictionSet(statisticsQuery.getBioEntityIdRestrictionSet());
// process OR conditions
Multiset<Integer> condGenes =
getScoresForOrConditions(orConditions, statisticsStorage, scoringExps);
if (results == null) results = condGenes;
else {
Iterator<Multiset.Entry<Integer>> resultGenes = results.entrySet().iterator();
while (resultGenes.hasNext()) {
Multiset.Entry<Integer> entry = resultGenes.next();
if (!condGenes.contains(
entry.getElement())) // AND operation between different top query conditions
resultGenes.remove();
else
// for all gene ids belonging to intersection of all conditions seen so far, we
// accumulate experiment counts
results.setCount(
entry.getElement(), entry.getCount() + condGenes.count(entry.getElement()));
}
}
}
}
if (results == null) {
results = HashMultiset.create();
}
return results;
}