4 drivers of success in drug repurposing for rare diseases

2023 marks the 40th anniversary of the Orphan Drug Act(opens in new tab/window) in the US, which catalyzed the development of hundreds of therapies for rare disorders, over half of which were approved in the last decade. Despite this progress, there are more than 300 million people worldwide living with a rare condition, half of whom are children, and only 5%(opens in new tab/window) of 8,000 known rare diseases have licensed treatments.

This is why Elsevier has launched its Year of the Zebra initiative(opens in new tab/window), aiming to improve education about rare diseases — patients of which are known as zebras in medical education as they are less common than horses. One strategy that’s emerged as a pragmatic and potentially life-changing solution to the rare disease challenge is drug repurposing — the exploration of new uses for approved, discontinued and investigational drugs.

A complex path

For drug repurposing, researchers take one of three approaches: drug-centric, disease-centric or target-centric. The most common method for rare diseases is disease-centric repurposing; this involves gaining an understanding of the underlying biological mechanisms of a disease, then identifying existing treatments that act on molecular targets involved in the mechanisms, such as interactions with specific proteins, enzymes or signaling pathways. Rare disease challenges that complicate this process include often poorly understood biology of conditions, small patient populations, and more stringent regulations when it comes to childhood medications.

Two use cases where these challenges were successfully mitigated to uncover drug repurposing candidates include diffuse intrinsic pontine glioma (DIPG) and Friedreich’s ataxia. Analyzing the similarities between them, we identify four driversto make rare disease repurposing projects successful.

Using AI to uncover therapies for childhood brain cancer

A collaborative project between Elsevier and the Sinergia Consortium(opens in new tab/window) brought together data and disease experts to create in silico molecular pathway models for DIPG, a rare form of childhood brain cancer. The complex and poorly understood maturation pathways of human neuron precursors had limited treatment discovery for DIPG. Researchers used AI techniques to better understand these pathways and identify drug repurposing candidates. Using Elsevier’s Biology Knowledge Graph, the team analyzed omics data from real-world DIPG patients to identify the most active genes in DIPG. They focused on TP53, a cancer suppressor, and one of the most frequently mutated genes in cancer cells that drives radio resistance.

The team text-mined PubMed abstracts and full-text journal articles for reports of drugs inhibiting “mutant TP53.” They compiled a list of 144 FDA-approved drugs known to target this mutation, scoring candidate drugs according to GSEA enrichment score(opens in new tab/window) and by the number of inhibited regulators. They also predicted an additional 322 drugs to inhibit mutant TP53 based on the mechanisms for known mutant TP53 inhibitors. The intersection of known and predicted mutant TP53 inhibitors with drugs predicted to inhibit DIPG disease model and with drugs predicted to inhibit DIPG tumor expression yielded a list of 284 drugs; the top-ranking drugs were vorinostat, aspirin, dasatinib, niclosamide and sirolimus.

Repurposing for Friedreich’s ataxia in under an hour

In a demonstration that took just 40 minutes, a SciBite and Stardog(opens in new tab/window) partnership uncovered drug repurposing candidates to treat Friedreich’s ataxia, a rare genetic disorder caused by a deficiency of frataxin — a small nuclear-encoded mitochondrial protein. A schema was built in SciBite’s ontology management platform CENtree(opens in new tab/window), and the ontologies were deployed within TERMite(opens in new tab/window), an entity recognition engine. TERMite automatically expanded synonyms and used these to identify entities within unstructured text. Entities were mapped back to a common identifier — turning unstructured documents into structured semantic data, rapidly extracting terms from 4,000 documents to create a knowledge graph.

The resulting knowledge graph connected data in a way that was not possible manually or in siloed systems. The graph collated data and metadata, enabling a simple way for users and machines to query the disease. It pinpointed four Phase IV drugs that increased the expression of frataxin, including tranylcypromine and etravirine.