SeMRA

The Semantic Mapping Reasoner and Assembler (SeMRA) is a Python package that provides:

1. An object model for semantic mappings (based on SSSOM)
2. Functionality for assembling and reasoning over semantic mappings at scale
3. A provenance model for automatically generated mappings
4. A confidence model granular at the curator-level, mapping set-level, and community feedback-level

We also provide the SeMRA Raw Semantic Mappings Database, a set of pre-assembled semantic mappings from hundreds of ontologies and databases, on Zenodo at https://doi.org/10.5281/zenodo.11082038 that can be rebuilt with semra build. More information here.

💪 Getting Started

Here's a demonstration of SeMRA's object, provenance, and cascading confidence model:

from semra import Reference, Mapping, EXACT_MATCH, SimpleEvidence, MappingSet, MANUAL_MAPPING

r1 = Reference(prefix="chebi", identifier="107635", name="2,3-diacetyloxybenzoic")
r2 = Reference(prefix="mesh", identifier="C011748", name="tosiben")

mapping = Mapping(
    subject=r1,
    predicate=EXACT_MATCH,
    object=r2,
    evidence=[
        SimpleEvidence(
            justification=MANUAL_MAPPING,
            confidence=0.99,
            author=Reference(
                prefix="orcid", identifier="0000-0003-4423-4370", name="Charles Tapley Hoyt"
            ),
            mapping_set=MappingSet(
                name="biomappings",
                license="https://creativecommons.org/publicdomain/zero/1.0/",
                confidence=0.90,
            ),
        )
    ],
)

Assembly

Mappings can be assembled from many source formats using I/O functions exposed through the top-level semra submodule:

import semra

# load mappings from any standardized SSSOM file as a file path or URL, via `pandas.read_csv`
sssom_url = "https://w3id.org/biopragmatics/biomappings/sssom/biomappings.sssom.tsv"
mappings = semra.from_sssom(
    sssom_url,
    license="spdx:CC0-1.0",
    mapping_set_title="biomappings",
)

# alternatively, metadata can be passed via a file/URL
mappings_alt = semra.from_sssom(
    sssom_url, metadata="https://w3id.org/biopragmatics/biomappings/sssom/biomappings.sssom.yml"
)

# load mappings from the Gene Ontology (via OBO format)
go_mappings = semra.from_pyobo("go")

# load mappings from the Uber Anatomy Ontology (via OWL format)
uberon_mappings = semra.from_pyobo("uberon")

SeMRA also implements custom importers in the semra.sources submodule. It's based on a pluggable architecture (via class-resovler) so additional custom sources can be incorporated without modifying the SeMRA source code.

from semra.sources import get_omim_gene_mappings

omim_gene_mappings = get_omim_gene_mappings()

Inference

SeMRA implements the chaining and inference rules described in the SSSOM specification. The first rule is inversions:

from semra import Mapping, EXACT_MATCH, Reference
from semra.inference import infer_reversible

r1 = Reference(prefix="chebi", identifier="107635", name="2,3-diacetyloxybenzoic")
r2 = Reference(prefix="mesh", identifier="C011748", name="tosiben")

mapping = Mapping(subject=r1, predicate=EXACT_MATCH, object=r2)

# includes the mesh -> exact match-> chebi mapping with full provenance
mappings = infer_reversible([mapping])

graph LR
    A[2,3-diacetyloxybenzoic<br/>chebi:107635] -- skos:exactMatch --> B[tosiben<br/>mesh:C011748]
    B -. "skos:exactMatch<br/>(inferred)" .-> A

The second rule is about transitivity. This means some predicates apply over chains. SeMRA further implements configuration for two-length chains and could be extended to arbitrary chains.

from semra import Reference, Mapping, EXACT_MATCH
from semra.inference import infer_chains

r1 = Reference.from_curie("mesh:C406527", name="R 115866")
r2 = Reference.from_curie("chebi:101854", name="talarozole")
r3 = Reference.from_curie("chembl.compound:CHEMBL459505", name="TALAROZOLE")

m1 = Mapping(subject=r1, predicate=EXACT_MATCH, object=r2)
m2 = Mapping(subject=r2, predicate=EXACT_MATCH, object=r3)

# infers r1 -> exact match -> r3
mappings = infer_chains([m1, m2])

graph LR
    A[R 115866<br/>mesh:C406527] -- skos:exactMatch --> B[talarozole<br/>chebi:101854]
    B -- skos:exactMatch --> C[TALAROZOLE<br/>chembl.compound:CHEMBL459505]
    A -. "skos:exactMatch<br/>(inferred)" .-> C

The third rule is generalization, which means that a more strict predicate can be relaxed to a less specific predicate, like owl:equivalentTo to skos:exactMatch.

from semra import Reference, Mapping, EXACT_MATCH
from semra.inference import infer_generalizations

r1 = Reference.from_curie("chebi:101854", name="talarozole")
r2 = Reference.from_curie("chembl.compound:CHEMBL459505", name="TALAROZOLE")

m1 = Mapping(subject=r1, predicate=EXACT_MATCH, object=r2)

mappings = infer_generalizations([m1])

graph LR
    A[talarozole<br/>chebi:101854] -- owl:equivalentTo --> B[TALAROZOLE<br/>chembl.compound:CHEMBL459505]
    A -. "skos:exactMatch<br/>(inferred)" .-> B

The third rule can actually be generalized to any kinds of mutation of one predicate to another, given some domain knowledge. For example, some resources curate oboInOwl:hasDbXref predicates when it's implied that they mean skos:exactMatch because the resource is curated in the OBO flat file format.

from semra import Reference, Mapping, DB_XREF
from semra.inference import infer_dbxref_mutations

r1 = Reference.from_curie("doid:0050577", name="cranioectodermal dysplasia")
r2 = Reference.from_curie("mesh:C562966", name="Cranioectodermal Dysplasia")
m1 = Mapping(subject=r1, predicate=DB_XREF, object=r2)

# we're 99% confident doid-mesh dbxrefs actually are exact matches
mappings = infer_dbxref_mutations([m1], {("doid", "mesh"): 0.99})

graph LR
    A[cranioectodermal dysplasia<br/>doid:0050577] -- oboInOwl:hasDbXref --> B[Cranioectodermal Dysplasia<br/>mesh:C562966]
    A -. "skos:exactMatch<br/>(inferred)" .-> B

Processing

Mappings can be processed, aggregated, and summarized using functions from the semra.api submodule:

from semra.api import filter_minimum_confidence, prioritize, project, summarize_prefixes

mappings = ...
mappings = filter_minimum_confidence(mappings, cutoff=0.7)

summary_df = summarize_prefixes(mappings)

# get one-to-one mappings between entities from the given prefixes
chebi_to_mesh = project(mappings, source_prefix="chebi", target_prefix="mesh")

# process the mappings using a graph algorithm that creates
# a "star" graph for every equivalent entity, where the center
# of the star is determined by the equivalent entity with the
# highest priority based on the given list
priority_mapping = prioritize(
    mappings,
    priority=[
        "chebi",
        "chembl.compound",
        "pubchem.compound",
        "drugbank",
    ],
)

The prioritization described by the code above works like this:

graph LR
    subgraph unprocessed [Exact Matches Graph]
    A[R 115866<br/>mesh:C406527] --- B[talarozole<br/>chebi:101854]
    B --- C[TALAROZOLE<br/>chembl.compound:CHEMBL459505]
    A --- C
    end
    subgraph star [Prioritized Mapping Graph]
    D[R 115866<br/>mesh:C406527] --> E[talarozole<br/>chebi:101854]
    F[TALAROZOLE<br/>chembl.compound:CHEMBL459505] --> E
    end
    unprocessed --> star

🏞️ Landscape Analysis

We demonstrate using SeMRA to assess the landscape of five biomedical entity types:

1. Disease
2. Cell & Cell Line
3. Anatomy
4. Protein Complex
5. Gene

These analyses are based on declarative configurations for sources, processing rules, and inference rules that can be found in the semra.landscape module of the source code. These can be rebuilt with semra landscape, with more documentation here.

🤖 Tools for Data Scientists

SeMRA provides tools for data scientists to standardize references using semantic mappings.

For example, the drug indications table in ChEMBL contains a variety of references to EFO, MONDO, DOID, and other controlled vocabularies (described in detail in this blog post). Using SeMRA's pre-constructed disease and phenotype prioritization mapping, these references can be standardized in a deterministic and principled way.

import chembl_downloader
import semra
from semra.api import prioritize_df

# A dataframe of indication-disease pairs, where the
# "efo_id" column is actually an arbitrary disease or phenotype query
df = chembl_downloader.query("SELECT DISTINCT drugind_id, efo_id FROM DRUG_INDICATION")

# a pre-calculated prioritization of diseases and phenotypes from MONDO, DOID,
# HPO, ICD, GARD, and more.
url = "https://zenodo.org/records/15164180/files/priority.sssom.tsv?download=1"
mappings = semra.from_sssom(url)

# the dataframe will now have a new column with standardized references
prioritize_df(mappings, df, column="efo_id", target_column="priority_indication_curie")

🚀 Installation

The most recent release can be installed from PyPI with uv:

$ uv pip install semra

or with pip:

$ python3 -m pip install semra

The most recent code and data can be installed directly from GitHub with uv:

$ uv pip install git+https://github.com/biopragmatics/semra.git

or with pip:

$ python3 -m pip install git+https://github.com/biopragmatics/semra.git

👐 Contributing

Contributions, whether filing an issue, making a pull request, or forking, are appreciated. See CONTRIBUTING.md for more information on getting involved.

👋 Attribution

⚖️ License

The code in this package is licensed under the MIT License.

📖 Citation

Assembly and reasoning over semantic mappings at scale for biomedical data integration
Hoyt, C. T., Karis K., and Gyori, B. M.
Bioinformatics, 2025, btaf542.

@article{hoyt2025semra,
    author = {Hoyt, Charles Tapley and Karis, Klas and Gyori, Benjamin M},
    title = {Assembly and reasoning over semantic mappings at scale for biomedical data integration},
    journal = {Bioinformatics},
    pages = {btaf542},
    year = {2025},
    month = {09},
    issn = {1367-4811},
    doi = {10.1093/bioinformatics/btaf542},
    url = {https://doi.org/10.1093/bioinformatics/btaf542},
    eprint = {https://academic.oup.com/bioinformatics/advance-article-pdf/doi/10.1093/bioinformatics/btaf542/64428329/btaf542.pdf},
}

🍪 Cookiecutter

This package was created with @audreyfeldroy's cookiecutter package using @cthoyt's cookiecutter-snekpack template.

🛠️ For Developers

$ git clone git+https://github.com/biopragmatics/semra.git
$ cd semra
$ uv pip install -e .

$ python3 -m pip install -e .

$ uvx pre-commit install

$ pip install pre-commit
$ pre-commit install

$ tox -e py

$ git clone git+https://github.com/biopragmatics/semra.git
$ cd semra
$ tox -e docs
$ open docs/build/html/index.html

🧑‍💻 For Maintainers

$ uv tool install keyring
$ keyring set https://upload.pypi.org/legacy/ __token__
$ keyring set https://test.pypi.org/legacy/ __token__

$ tox -e finish

$ cruft update