Please login to the form below

Not currently logged in

AI and genomics: a revolution in drug discovery and development

Artificial intelligence can make drug discovery and development faster and less risky according to an industry expert, who said understanding complex diseases at the genetic level is now a possibility.

Data has always been key to drug development. Pharmaceutical products are approved based on empirical evidence of safety and efficacy that is generated in preclinical experiments and clinical trials.

Data is also core to the discovery process, with developers selecting candidates that have shown through experimentation that they interact with their chosen disease targets.

In recent years, the way data is used for drug discovery and development has changed. The need to refill pipelines in a cost-efficient manner forced innovative pharma to look for faster, more efficient ways of identifying drug candidates.

Pfizer, for example, has used a machine learning system called IBM Watson to search for cancer drug targets. Likewise, Sanofi has partnered with Exscientia to use its artificial intelligence technology to find potentials targets for metabolic disease therapies.

AI innovation

But artificial intelligence can do more than just find targets, according to Dr Steve Gardner from Precisionlife, who said a wider, ‘multi-omic’ approach aligned with modern R&D principles can help elucidate the complex genetics underlying diseases.

“Drug discovery involves looking at the role genes play in disease, by examining patient phenotype. While this approach has merit, diseases usually involve multiple genes. Determining how these multiple genes interact is challenging as current ‘omics’ techniques lack the required resolution.”

To address this, Precisionlife – known until November 2019 as Row Analytics – developed a proprietary framework that improves the resolution and accelerates the search for diseaserelevant genes.

Gardner continued: “We are focusing the search on SNPs predicted to be most likely to add to an existing combination of genes – the signature – in a way that increases its biological significance.

“Once we have identified all the SNPs associated with a disease we can use this to stratify different patient subgroups who share different combinations of those SNPs.

“For a given patient subgroup we can then identify which signatures are associated with their form of disease. We can then see which genes correspond to the SNPs we have identified and look for the biological relationships between genes involved in a particular disease state.”

Research strategy

Precisionlife’s system is designed around the so-called 5Rs principles of research effectiveness developed by AstraZeneca. In 2011, AZ looked at its R&D activities between 2005 and 2010 and identified the characteristics of successful projects.

These characteristics helped AZ develop an R&D strategy in which every project seeks to establish the right target, right tissue, right safety, right patient and right commercial potential – the 5Rs.

Last year, AstraZeneca said the approach helped it achieve a five-fold improvement in R&D success – defined as the percentage of compounds taken from the laboratory through to phase 3 clinical trials.

Using the 5Rs approach with Precisionlife’s framework offers developers a powerful discovery tool, according to Gardner, who cited an asthma study the firm undertook using data from 35,000 patients in the UK Biobank as an example.

“There are several different subpopulations of asthma sufferers who respond differently to drugs, and it was suspected they have different underlying genetics. We focused on the TH2+ and TH2- populations and set out to find the combinations of genes responsible for both types of disease to see how similar or different they were.

“We analysed the SNPs and genes associated with disease risk in both populations and compared the two sets. They turned out to be almost completely different from each other – indicating two different diseases that happen to share the common symptoms of breathlessness, wheezing,” he said.

Gardner’s team then annotated the new genes, finding out as much as possible about every mutation before moving to the next stage.

“We selected the most promising targets and took them through an automated in silico validation process. Our information included details about the tissue expression, chemistry safety and bioavailability and the ‘druggability’ of each gene target.”

The next step of the process was to work with a partner to run the most promising targets through various confirmatory assays, including phenotypic assessments using stem cells from patients.

Gardner continued: “Our approach allowed us to identify and validate (insilico) the most promising targets within a single day and to prioritise and run multiple targets through phenotypic assays in two to three months, which represents a substantial timesaving over conventional approaches.”

Research DNA

The potential of the combined omics/AI is clear. But how readily the pharmaceutical industry will adopt the approach for drug discovery is a topic of debate.

At present drug discovery relies on established screening methods where the target or a candidate molecule is the focus. Convincing drug firms to switch to a sophisticated, data analytics based screening approach is likely to take time, Gardner said.

“Our approach is significantly enabling for genomic medicine and drug discovery. The potential to rapidly analyse large data sets and quickly find genes and combinations of genes associated with identifiable patient subgroups that are potential targets for drug development could have a dramatic impact on industry R&D.

“But the pharmaceutical industry is understandably cautious and we expect a degree of scepticism. However, to date we have done 15 validation studies. The quickest way to convince drug developers of the innovation and benefits of our approach has been to show them.”

Gardner told us mid-sized drug companies are likely to be most receptive to discovery innovation, explaining that “they have to innovate – as do biotech companies – because they need to find new targets and more effective ways of carrying out R&D”.

The omics approach could also help biotech companies attract acquisitive biopharma companies. Traditional biotech backers usually aim to channel $10m into projects to support development of a single candidate over a period of several years until it can be sold to a larger developer.

Gardner continued: “By being able to access well-validated innovation in targets for their chosen diseases, biotechs de-risk the development process and may be able to bring more programmes to a stage where they can be licensed to pharma.”

Other purposes

The approach can be applied later in the drug development process. In clinical research, for example, the strategy can be used to examine patient data sets to identify targets and guide study design, Gardner added.

“In clinical trials recruitment is a challenging and costly process. We can help to stack the odds in the sponsors’ favour by identifying ‘signatures’ that can be used to quickly and inexpensively find patients belonging to particular subgroups.

“We can also generate new findings from old data sets. For example, we looked at genetic data from patients with ALS. Previous research had identified 35 genes that play a role in the disease. Using our approach we quickly identified and validated with multiple KOLs 33 additional genes that were strongly associated with ALS.”

Ten of the new ALS-related genes are being tested using an existing neural progenitor cell assay at the Sheffield Institute for Translational Neuroscience (SITraN).

Drug repurposing – the identification of new indications for approved products – is another potential application. Again it is Precisionlife’s ability to identify gene combinations, and from them patient subgroups, that is key to such uses.

Gardner concluded: “Many pharmaceutical companies are looking at repurposing: does an approved drug have potential as a treatment for a new condition? Can we identify targets against which approved drugs are effective? Our framework can help answer these questions.”

Gareth Macdonald is a freelance journalist specialising in the life sciences industry

Article by
Gareth Macdonald

4th June 2020

Article by
Gareth Macdonald

4th June 2020

From: Research


Career advice

No results were found

Subscribe to our email news alerts


Add my company
Springer Nature

Operating in 50+ countries, we’re a leading research, educational, and professional publisher dedicated to advancing scientific discovery. Whether you’re looking...