An v2 extension of the Beacon protocol will allow the query for additional
data beyond genome variants, using a proposed filters extension. Such filters
are thought to be prefixed attributes, where the (public or private) prefix
becomes the basis of scoping the value to the correct database value.
The Beacon v2 API supports the discovery of genomics and clinical datasets, and includes a powerful feature to enable the “filtering” of beacon responses by biomedical properties (e.g. phenotypes) and procedural metadata.
Filters belong to one of currently three super-classes:
Filters correspond to classes from bio-ontologies for biomedical data or
procedural metadata that are contained in public repositories such as the
Ontology Lookup Service or
NCBO’s BioPortal. Filters are
identified using the full bio-ontology term/class identifier as CURIE, e.g.
“HP:0100526”. Examples of bio-ontologies are, among others:
HP:0100526: Neoplasm of the lungPATO:0020002: female genetic sexNCIT:C8430: Ovarian Papillary TumorNCIT:C48724: T2 Stage FindingEFO:0009656: neoplastic sampleCustomFilters are biomedical or metadata terms that are locally defined by a
Beacon (e.g. not corresponding to known bio-ontology terms). As with standard
Filters, the Beacon v2 API is agnostic to how CustomFilters are implemented
by the Beacon, so this permits maximum flexibility with regards to identifying,
labelling and grouping CustomFilters. For example, related phenotype terms or
experimental sets could be grouped into local “dictionaries”, which could be
addressed through a local identifier. The only requirement is that each custom
filter term contains a unique identifier that can be used in Beacon requests.FuzzyFilters are implementations of classifiers which allow for some
alternatives in matching and mostly can be drop-ins where ontologies are
incomplete. Logically, through the potential matching of multiple values they
provide a limited alternative mechanism to allow OR-style queries.Beacon filter requests are simple yet flexible, and can be used to query qualitative or quantitative properties. For example, a qualitative phenotype can be represented by a single observation:
HP:0100526)filters=HP:0100526By default, the use of Filters in a Beacon request implies that a hierarchical ontology search is requested, whereby the Beacon will query for entities associated with the submitted term and all descendent terms.
Both qualitative and quantitative properties can be represented by attribute +
value pairs. Equality and relational operators (= < >) can be used
between attributes and values. Additionally, values can be associated with units
if applicable:
filters=HP:0100526,PATO:0020002,PATO:0000011>P70YSimilar query flexibility is shared by CustomFilters, where attributes and value pairs can be combined, for example:
customFilters=mydict.aterm:avalue,mydict.aterm2:avalue2customFilters=demographic.ethnicity:asian,demographic.sex:maleIn the examples above, filters are separated with commas.
The use of a filter term can be ambiguous if the entity to which it applies is not specified. For example, the term “metastatic melanoma” (EFO:0002617) could refer to an individual with metastatic melanoma, or a metastatic melanoma biosample.
The entity to which the filter applies can be forced by declaring the scope as a prefix in dot-annotation, for example:
filters=individual.EFO:0002617filters=biosample.EFO:0002617Currently a the logical AND is implied between filters. A limited way of OR
type queries can be provided through fuzzy FuzzyFilters.
All filtering terms used in a Beacon can be listed in order to show the range of biomedical and metadata properties described by the Beacon, and to assist with building requests.
An unordered unique value list containing all Filters and CustomFilters is
returned from the /filtering_terms endpoint. For each term, the following
information is returned:
source name
id
label
An earlier discussion had proposed a scoped query object design, following the originally proposed GA4GH schema’s object model.
