How Knowledge Graphs Are Accelerating Drug Repurposing and Indication Expansion

Drug discovery has always been expensive and uncertain. But the modern pipeline has made a difficult process almost unsustainably hard. Bringing a single drug to market can cost over $2 billion and take more than 10 years, yet nearly 90% of clinical candidates fail before approval. Unexpected toxicities, limited clinical efficacy, and ever-escalating R&D budgets cause the majority of candidates to fail.

Much of this failure traces back to one root cause: the evidence needed to make good decisions exists but it's scattered, siloed, and impossible to connect. Researchers making high-stakes decisions about which targets to pursue, which compounds to advance, and which combinations to test are doing so with an incomplete picture.

The problem isn't a lack of data. It's a lack of integrated, queryable data assembled in a way that reflects how biology actually works.

Indication expansion and drug repurposing have emerged as a strategic response: instead of starting from scratch, identify new therapeutic applications for assets that already have safety, pharmacokinetic, and manufacturing data behind them. The potential to compress timelines and reduce translational risk is enormous. The challenge is doing it systematically, at scale, with scientific evidence.

Knowledge graphs as the connective tissue of R&D

The core insight behind modern biomedical knowledge graph platforms is simple: biology is a network. Genes, proteins, pathways, diseases, drugs, and clinical outcomes don't exist in isolation -they interact in complex, non-linear ways. Understanding those interactions requires infrastructure that can represent and reason about relationships, not just store tables of data.

A well-built knowledge graph platform integrates public datasets, proprietary internal data, and scientific literature into a unified, AI-ready framework. Queries that would have required months of manual curation can be answered in days. Hypotheses that would never have surfaced through traditional analysis become visible through network-based reasoning. This is exactly what Elucidata’s Polly KG framework has been doing.

Elucidata’s knowledge graph’s computational approaches for indication expansion involves:

Network proximity analysis: Measures topological distance between drug targets and disease modules. Closer proximity = stronger biological rationale for repurposing.
Subgraph similarity: Identifies structural resemblance between biological pathways, surfacing repositioning candidates even when primary targets don't match.
Graph neural networks: Deep learning models that operate directly on the graph, predicting links and inferring relationships that haven't been explicitly documented.

The architecture is typically tiered: a foundational base layer drawing from broad public biomedical sources, a second layer of indication-specific context from curated oncology, immunology, or disease repositories, and a top layer incorporating proprietary internal data. This hierarchy allows you to build on shared knowledge while keeping their competitive assets protected and integrated.

Case Study: Finding a combination therapy for drug resistance in oncology

A pharmaceutical organization was studying acquired resistance to a targeted kinase inhibitor. Tumor cells were bypassing the therapy by activating compensatory signaling pathways. The evidence needed to understand these mechanisms was fragmented across functional screening data, pharmacovigilance records, public oncology repositories, and unstructured literature with no connection.

The organization partnered with Elucidata and structured its oncology datasets- bulk and single-cell RNA-seq, functional screening outputs, clinical trial records, real-world evidence, and internal research reports into a unified knowledge graph. What had been a collection of disconnected signals became a queryable biological network.

Outcome

Polly KG enabled the R&D team to generate a highly prioritized and biologically de-risked combination therapy hypothesis for experimental validation. The platform identified Drug A + Drug B as a promising strategy to overcome targeted therapy resistance, with a strong predicted bliss synergy score. This enabled the organization to accelerate in vitro validation while extending the therapeutic potential and commercial value of a flagship oncology asset

Conclusion

Drug repurposing and indication expansion aren't new ideas. What's new is the ability to pursue them systematically with the full depth of available biological and clinical evidence, integrated and queryable in real time. Knowledge graph platforms are becoming the connective tissue that makes this possible.

For organizations sitting with unexplored potential compounds, the question is no longer whether this kind of infrastructure is worth building. It's how quickly they can put it to work.

Explore how knowledge graph–driven indication expansion can accelerate your R&D pipeline.

If you’re interested more on how knowledge graphs are applied in real-world research explore these studies:

‍