Jonathan Dry, principal scientist, bioinformatics, oncology, innovative medicines & early development at AstraZeneca, explains how his company is using new technologies to simulate how drugs affect cancer pathways without the need for a trial

What is your background and current role? 

My training is in biomedical sciences (specialising in disease genetics) and Bioinformatics. All of my career to date has been spent in the pharmaceutical industry, focused on drug discovery and translational science for oncology. I have built a particular specialty in transcriptomics (analysis of mRNA changes associated to disease and drug response, and interpretation of their biological meaning) and more recently, next generation sequencing to identify genetic mutations driving cancer. 

What does your work involve? 

I manage a team of 12 supporting AstraZeneca's Oncology portfolio from target discovery through to early clinical trials. Typical projects may involve: 

  • Analysis of 'omic data from a particular patient cohort to understand variability and what might be the driving process that we can target with a new drug 
  • Analyses comparing drug response in pre-clinical models to molecular and phenotypic data to understand the biomarker differences between responding and non-responding patients, and resistance mechanisms that may be overcome with new drugs 
  • Development of algorithms to read data from clinical diagnostic assays to assess the likelihood of a patient responding to specific drugs. 

What are the challenges in this field? 

For bioinformatics in cancer we are constantly challenged with data of increasing depth and breadth. In the last five years, we've seen major achievements in the field of next generation sequencing that explores disease genetics at the patient level. New technologies are now emerging that can retrieve this information from each individual tumour cell and surrounding microenvironment and immune system, at multiple time points. These technologies complement a change in focus from tumour cell genetics to the environment in which that tumour exists – considering the influence of its surrounding microenvironment, immunology and patient health. The potential for new advances in computation to harness these data, potentially even simulate the disease, and assist our understanding of how best to therapeutically intervene is phenomenal and essential. 

What do you consider the biggest recent advance?

In oncology bioinformatics, this would have to be the toolkits developed to harness information in next generation sequencing data to understand the genetic drivers and diversity of cancer types, leading to a new age of molecularly targeted therapies now benefiting patients. 

Can you tell us a little about AstraZeneca's collaboration with Microsoft and BMA technology? 

AstraZeneca encourages an open research environment where our scientists can freely share their ideas and collaborate on projects with the best external partners. Our partnership with Microsoft is a great example of this. Together we have demonstrated something we believe represents a groundbreaking advance in drug discovery methodology. The Bio Model Analyzer (BMA) is a cloud-based simulation tool that allows scientists to model key signalling pathways in cancer cells in a more systematic way than we could in the past. To apply BMA in our projects, AstraZeneca provided the scientific know-how while Microsoft developed the sophisticated algorithms used to model variations in signalling pathways. The BMA works in a simple, logical way. It captures the way a biologist would think about the pathway instead of trying to fully mathematically model it, which makes it easy to use. 

What is 'drag and drop' cancer simulation? 

Using the BMA, our drug discovery scientists can capture and bring to life the basic signalling pathways that can go wrong in cancer cells by simply 'dragging and dropping' cells, genes and proteins on their computer screens and then drawing in the interactions between them. Once that's done, they can simulate lab experiments by mimicking, say, a drug inhibition at a certain point in that pathway – the computers do all the calculations to fill in the gaps and predict what happens if different steps are blocked with drugs. 

We can model the way many different proteins interact, simulate experiments that we previously did in the lab and test the likely effects of our drugs at different places on the pathway in order to identify the most promising places to intervene. 

What type of cancer are you focusing on and what have you discovered? 

Our initial focus has been on a type of blood cancer called acute myeloid leukaemia (AML). AstraZeneca scientists collected various types of data from AML cell lines in our research lab. At the same time, Microsoft computer scientists collated published knowledge and developed algorithms to model all the possible variations in a key signalling pathway. We then simulated experiments to predict how drug combinations might be used to overcome resistance to drugs that target cell signalling in patients with AML. Even I was surprised by how accurate the simulation was. It told us which drug combination would make a cell sensitive to treatment and the protein changes that led to that cell becoming sensitive. 

What are the key benefits of using computer modelling in research? 

Experiments that would previously have been done in the lab can be simulated by computers at far greater volume and speed. There could be more than a hundred different possible points to target with a drug in a single pathway. Faced with a so many choices, scientists tend to choose their favourites to investigate and may miss better options. The beauty of computerised modelling is that there's no limit on the number of hypotheses you can investigate, which removes both the guesswork and the potential for human bias. 

What does all this mean for patients?

Instead of hitting a pathway just because we know it's active, we can identify the best place to hit the pathway in order to get the best possible response for that patient. It also has the potential to move personalised medicine beyond its current boundaries. Current treatments target single points in signalling pathways, based on genetic mutations in a patient's tumour. But future personalised therapy is likely to focus on multiple points
in the pathway initially identified by computer simulations and confirmed in the laboratory. Ultimately, such simulations could even take account of interactions between the tumour and its broader environment, including host immunity and DNA damage response mechanisms. 

Do you foresee an increase in this approach to drug targeting? 

Yes. As more knowledge about different pathways and different cells is captured in these models, each an individual jigsaw piece, this then opens up the possibility of fitting the pieces together to model the cancer as a whole. As a result, the accuracy of these models will continue to improve and their value and utility continue to grow. 

Aside from discovering new drug targets, are there any other potential health applications for such computer simulation tools? 

I would say that precision medicine (understanding the best therapy for an individual patient) is a primary area of focus, but really any biological tissue and process could be modelled this way. For example models could also simulate the effect of a drug on normal tissues and organs influencing drug side effects or broader patient health and wellbeing. 

What keeps you awake at night? 

A new age of truly big data and real-time health management is upon us. My own specialty field of disease genomics has expanded to a previously unfathomable scale, but will potentially be eclipsed with the real time biological data collection possible from new biosensor technologies such as wearables and attempts to interconnect all this disparate data and information for a patient. The possibilities are endless. 