Skip to content

Mira documentation

MIRA — Multi-source Indication and Report Analytics

Open Targets Mira (Multi-source Indication and Report Analytics) harmonises clinical trials, regulatory records, drug indications, and safety evidence into consistent datasets for drug-discovery analysis.

The documentation is being developed around three questions:

  1. What are a Clinical Report and a Clinical Indication?
  2. How does Open Targets generate them with Mira's default workflow?
  3. How can a user change the YAML configuration for a different use case?

Tip

Start with How Mira works for the complete path from provider data to Clinical Reports, mapped entities, and Clinical Indications.

Installation

Install the package from PyPI with its NER extra:

pip install "opentargets-mira[ner]"

The ner extra installs the named-entity recognition models (PyTorch and Transformers) that drug and disease mapping, the EMA and PMDA providers and the mira CLI need. Without it, only the parts of Mira that do not touch NER can be imported.

The legacy private ChEMBL Oracle curation utilities require the optional Oracle extra:

pip install "opentargets-mira[oracle]"

The Oracle curation process is not required for the public default workflow and is intended for deprecation.

Data documentation

  • How Mira works

    Follow the complete path from provider data to mapped Clinical Reports and Clinical Indications.

    Read the overview

  • Clinical Report

    Understand one traceable evidence record, its clinical stage, source, provider, and core fields.

    Explore Clinical Reports

  • Clinical Indication

    See how Mira groups reports into drug–disease relationships and derives their maximum clinical stage.

    Explore Clinical Indications

  • Entity mapping

    Learn how source labels become ChEMBL and EFO identifiers, including the NER fallback paths.

    Follow entity mapping

Providers

Each provider page explains what that source contributes and the selection rules applied before its records enter the shared Clinical Report dataset.

Using Mira

  • Generate with Python

    Call provider, mapping, and aggregation functions directly to make the data flow visible.

    Use the Python API

  • Run recipes

    Use the CLI and YAML recipes for repeatable runs and deployment-specific configuration.

    Configure the CLI

  • Prepare the inputs

    Find the upstream releases, database tables, file formats, and expected local layout.

    Review data inputs