Could machine learning present a paradigm shift in drug discovery?

Angharad Baldwin catches up with Dr Raymond Barlow, CEO at e-therapeutics, about what network-driven drug discovery can offer the pharmaceutical industry in its pursuit of new medicines

Dr Raymond Barlow, CEO, e-therapeutics

e-therapeutics, based in Oxford, UK, aims to be an industrial platform to aid the development of new and safer drugs.

The company takes a specialist approach to network biology and uses a computer-based drug discovery model.

Says Dr Barlow: “It’s a fascinating and exciting time for bioscience. Following the considerable advancement of computational biology, true innovation is now possible. The confluence of a number of different disciplines and a deeper understanding of immunology and cellular biology has led to the ability to compile, interrogate and use data in a more intelligent way, taking advantage of computational power.”

The ability to make use of sophisticated algorithms has given scientists, he says: “the capability to produce actionable output from collected data,” remarking that this is a “paradigm shift in the history of the pharmaceutical industry.”

Traditionally in patients, medicines are used to correct part of a biochemical pathway; but, quite often, their body will find a work around, leading to medicines not effectively treating disease. The fundamental premise of e-therapeutics is to have a more holistic understanding of the underlying biology underpinning disease and produce medications with greater efficacy.

This is achieved using computational techniques, data interrogation/augmentation and machine learning to test biological networks to find out which pathways are most susceptible to medical intervention ... and provide this information to pharmaceutical companies.

e-therapeutics has access to large databases of biological information, including complete maps of protein-protein interactions from 14 million small molecule compounds.

Accessing public domain databases, proprietary information and their own-curated database, this data is combined to produce an interactome. This network model of disease looks at a variety of different outcomes in both healthy and diseased tissues from many different sources, enabling the company to interrogate the data to identify medicines.

After this initial in silico research, in vitro and in vivo analysis can be performed to test the computational predictions.

When asked about the challenges faced, Dr Barlow comments: “Currently, the company is ahead of its time and it’s challenging to get the pharmaceutical industry to buy into a different way of thinking about drug discovery and to overcome ingrained perspectives.”

He continues: “The approach that a company has a black box and, therefore, you should trust me because the output is relevant and will work for you will never hold any water in the pharmaceutical industry. The providence of the data and the way in which you interrogate the testability of the output from the in silica platform is extremely important. Validation in real models of disease is vital in terms of progression.”

Dr Barlow also believes this data is not only useful from a drug discovery viewpoint, but that it can also be used to inform decisions right across the value chain — using this intelligence to drive clinical trial recruitment, clinical trial design and patient monitoring. He predicts the use of informatics in this way is “only going to get more and more powerful, and more and more influential.”

Pulling together a number of technological approaches, e-therapeutics has created a pragmatic drug discovery engine to accelerate research. They’re taking a computational hypothesis to a promising compound in a matter of months, as opposed to what has historically taken years.

Companies