Can better knowledge engineering save pharmaceutical R&D?

RAIR Health
RAIR Health
Published in
4 min readJun 11, 2021


Pharmaceutical research has, by most standards, been slow to fully embrace data technologies. Reasons for this include heavy investment in empirical “wet” science; regulatory-driven product development that is not easily disrupted; and a “not invented here” culture. But pharma researchers may have a point: big bets in new technology often failed to improve drug development while adding to costs. New technologies also create data that no longer sits neatly in columns and rows. This means deep analysis, like trying to identify subtle relationships between gene variation and disease, is difficult at best and far beyond what relational databases can do.

The need for more effective data technologies has never been stronger. Recent history shows that drug development is taking longer and becoming more expensive. The cost of developing a drug averages $2.6bn and failure rates are around 90%. The opportunity cost of each additional day of research is estimated at between $600k and $8m. Consequently, R&D returns are now below the cost of capital. With these challenges in mind RAIR Health started working on its knowledge engineering solution for pharma R&D.

Introducing Knowledge Graphs

Let’s start with a question: what if we could bring the world’s life science research together and link the data so you could find and understand a relationship between any two pieces of data, no matter their distance? What if we could do this without supercomputers or complicated machine learning algorithms? As you probably guessed, this is already possible with a type of AI technology called knowledge graphing.

Knowledge graphing is not new, so why is it making headlines now? It is because recent advancements (described below) have radically improved its usefulness, particularly in life sciences. At their core, all knowledge graphs employ a form of applied AI that integrates a semantic schema with multi-dimensional data, and they can do this at considerable scale. Early versions showed promise and gained popularity when companies like Facebook, Google and LinkedIn started to organize huge data repository using knowledge graphs. Three recent advances have now considerably extended its usefulness. First, entity and relationship abstractions now allow n-level hierarchies that allow greater sophistication of data structure. Second, relationships can take attributes and sub-relationships, which give greater connectedness to the data. And third, the introduction of a semantically-rich query language enables users to find logical inferences using fairly simple syntax. Taken together, these advancements have improved data insights and reduced time spent data wrangling, all without time consuming trial and error.

Knowledge Graphs in Pharma R&D

Knowledge graphing’s ability to identify relationships between disparate data is perfectly suited to pharma’s complicated data structures and multiple sector disciplines which are fully understood by precisely nobody! But despite their potential, knowledge graphs are still rare outside large companies. Companies like Roche, Novartis and AstraZeneca have in-house teams, and a few health tech start-ups are entering the market — Benevolent AI collaborating with Lilly on COVID-19 and with AstraZeneca on chronic kidney disease, for example — but small and mid-sized research companies rarely have the resources.

To build a knowledge graph team requires hiring expensive, and some might consider non-core, tech resources at a time when the industry has spent a decade getting out of non-core activities. Add to this the time it takes to collect data, often made difficult because of the complexity of accessing some data, and it can take months or even years to negotiate access to some sensitive data. So if a company wants to work with knowledge graphs, they have two choices: build in-house if budget and time allow or seek a “knowledge as a service” partner.

RAIR Knowledge Graph — Powerful Knowledge as a Service

Against this backdrop RAIR launched its own pharmaceutical knowledge graph called “RKG” as part of its Oculair.KE solutions. Comprising over 1.4 million entities and tens of millions of relationships, RKG uses state-of-the-art graph technology to connect human proteins, genes, pathways, SNPs, drugs and diseases. This gives users an opportunity to discover, interact with, and draw insights from a vast store of structured and unstructured life science data. While RAIR’s core business is focussed on ophthalmology, RKG includes the full range of human proteins, genes and diseases in reflection of the fact that a wide variety of common systemic conditions, such as Alzheimer’s disease, chronic kidney disease, and rheumatological and cardiovascular diseases, can be found in the structures of the eye.

RAIR has also negotiated a collaboration with a NHS Trust to procure anonymized clinical insights from real-world data. This means RAIR can offer RKG with a clinical knowledge module. It can also integrate client’s own data, creating a bespoke knowledge graph built to client specifications.

Because the time and cost of building the knowledge graph are sunk, RAIR’s solution can provide customers with cost-effective access to knowledge graphing without upfront investment.

Applications in Pharma R&D

RAIR’s knowledge graph is ideal for helping improve the process of drug discovery without fundamentally changing it. It can be implemented either as an ongoing “knowledge as a service” or as a discrete consultancy project with no added headcount or upfront cost. Besides drug discovery, RKG can also help reposition old drugs; identify biomarkers for patients who might respond to therapy or develop drug sensitivity; find new treatment pathways for more efficient healthcare; and find diagnostic markers for early detection of diseases. RAIR’s aim is to deliver a transformational knowledge tool at a cost which has a demonstrably positive return on investment. The ability to implement these modern adjacencies to the core business of drug discovery, together with improved productivity which flows from new insights, may do just that.

What can RAIR’s knowledge graph do for your research team? Check out or get in touch with us today to set up a RKG demonstration.