Optimising Drug Discovery R&D with Graphs

Data in Life Sciences expert Dr Alexander Jarasch on a new wave of advanced data handling techniques that point the way to new drug discovery breakthroughs

Notoriously, drug discovery is a difficult and complex area. 90% of clinical drug development fails and only 6% of all the drugs that go to clinical trial make it onto the market.

But could a new approach to assessing more promising pathways help make drug discovery a more tractable problem? The new approach uses the data modelling technique of graphs, as instantiated in a knowledge graph.

Knowledge graphs deliver value by working on the basis that every dataset is a connected element. Unlike traditional SQL databases, which store data in tables with fixed columns and rows, knowledge graphs store data as nodes (or entities) connected by edges (or relationships). It is in the power of those interconnections that the breakthrough insights lie. For example, in the Panama Papers journalism expose, utilising knowledge graphs made it possible to capture what was otherwise an impenetrably obscure network of offshore accounts, shell companies, and individuals hiding money.

The more interrelationships can be analysed, the richer the knowledge

Knowledge graphs are designed to represent complex data and can be used in a wide range of applications beyond financial investigations. For pharma, they can be used to represent the complex interrelationships and correlations between information about diseases, genes, the environment, diet, behaviour, and other factors.

In particular, graph databases have made it possible to perform very high-scale cross-comparisons involving billions of connections. These can help researchers identify patterns and connections that might not be immediately obviousand which has the potential to transform fields such as testing new candidates for drugs.

From 1% success rates to more like 15%

Let’s consider Servier's use case. Servier is a French-headquartered international pharmaceutical company specializing in oncology, neuroscience, immuno-inflammation, cardiometabolic and venous diseases, and generics. It is a multi-billion dollar company dedicated to being an agile, digital performer.

The company’s dedicated R&D function, the Institut de Recherches Servier, has long experience researching tricky small molecules. This pursuit is inherently difficult, as the vast majority of drug candidates are composed of relatively 'small' molecules, making it highly challenging to discern which ones warrant further exploration or not.

Small molecules, defined in chemistry as compounds with a molecular mass under 1,000 atomic mass units and smaller than proteins and nucleic acids, exert their effects by binding to and altering the functions of proteins and nucleic acids.

Given the diminutive size and comparable structural characteristics of the small molecule, discriminating between promising and futile search paths can be tricky. Pre-graph technology, the team would only find search candidates at the rate of less than 1%.

But with the new knowledge graph-based approach, Servier has achieved a consistent hit rate of 15% across a focused dataset—1,000 small molecules instead of anunwieldy 1 million. Graphs are a powerful way to capture even the most complex reality as a set of ‘nodes’ or entities that contain information as attributes and sit in a network of connections with their siblings.

The new knowledge graph-based decision support tool, Pegasus, efficiently sifts therapeutic targets to identify the most relevant screening modalities, helping Servier design more appropriate experiments.

This follows the successful creation of a library of small molecules and their relationships based on interactions with the graph database. It’s in this library that Servier identifies good small molecules by ingesting a wide range of heterogeneous information from pre-existing data, including the open-source medical knowledge banks.

As a result, Servier is convinced that knowledge graph technologies have the potential to revolutionise drug discovery and development by better organising complex data and generating knowledge to more accurately support decision-making.

When considering a complex problem space, we need to be efficient at choosing what to explore and what to ignore. This is true for all kinds of data analysis, but this is especially true in drug discovery. Based on the impressive results in starting to optimise the drug discovery process, Servier may have pioneered a truly groundbreaking approach. It may also be one that could be of use to many others.

Topics: AI & Digital