Data Science & Artificial Intelligence:

Unlocking new science insights

Data science and AI has the potential to transform the way we discover and develop new medicines – turning yesterday’s science fiction into today’s reality with the aim of enabling the translation of innovative science into life-changing medicines

Jim Weatherall Vice President, Data Science & AI, BioPharmaceuticals R&D

We are using AI to help us analyse and interpret huge quantities of data at all stages of drug discovery and development with the aim of:

 

  • Gaining a better understanding of the diseases we want to treat
  • Identifying new targets for novel medicines
  • Driving personalised medicine strategies
  • Speeding up the way we design, develop and make new drugs

 


Building disease understanding through knowledge graphs




If you’ve ever asked Google or Alexa a question, you will have tapped into knowledge graphs.

These bring information together from thousands of different sources to find you the answer you need.

At AstraZeneca, we’re using knowledge graphs to give our scientists the information they need about genes, proteins, diseases and compounds and how they relate to each other. Using AI to combine information from multiple sources we hope to draw more accurate conclusions than if we analysed science literature by hand. AI also has the potential to find patterns in these graphs revealing previously unexplored hypotheses.

Our knowledge graphs integrate genomic, disease, drug and safety information, so our researchers can ask key questions to help identify and prioritise drug targets. As our data and knowledge evolves so will our graphs, so every new experiment can benefit from everything learned before.

Ultimately, we want to develop personalised knowledge graphs that bring the right information to the right scientist, at the right time.




Predicting what molecules to make next and how to make them


Through AI, we have the potential to transform medicinal chemistry, augmenting traditional design with sophisticated computational methods to predict what molecules to make next and how to make them.

Garry Pairaudeau Head of Hit Discovery, Discovery Sciences, R&D



We are exploring the use of AI to help us discover new medicines. We believe it is has great potential to increase the quality and reduce the time it takes to discover a potential drug candidate.

Discovering a potential drug molecule requires several years of detailed scientific research; synthesising and testing of thousands of molecules in order to achieve the right drug properties.

AI is enabling us to rapidly generate novel ideas for molecules to make and rank these ideas using predictions based on large data sets available to us.

Having identified promising molecules, the next step is to synthesise the molecules in the laboratory. AI is starting to help here too – the science of synthesis prediction is rapidly evolving and we will soon be able to use AI to help us deduce the best way to make a molecule in the shortest time.

We see AI as a key component in the chemistry lab of tomorrow – not only for discovering and making new drugs but for controlling automation to speed up the repeated cycles of generating, analysing and testing high-quality compounds.

Using AI for fast, accurate image analysis




Every week, our pathologists analyse hundreds of tissue samples from our research studies. They check them for disease and for biomarkers that may indicate patients most likely to respond to our medicines. It is very time consuming which is why we are training AI systems to assist pathologists in analysing samples accurately and more effortlessly. This has the potential to cut analysis time by over 30%.

For one of our AI systems, we implemented an approach inspired from how some self-driving cars understand their environment. We trained the AI system to score tumour cells and immune cells for a biomarker, called PD-L1, which has potential to help inform immunotherapy-based treatment decisions for bladder cancer.

Our AI system looks at thousands of images from tissue samples, methodically checking each one for PD-L1. It saves our pathologists time and is especially useful in difficult cases.



Building the right data backbone


Today we are generating and have access to more data than ever before. Data and analytics have the potential to transform our business, but the true value of scientific data can only be realised if it is “FAIR”, or Findable, Accessible, Interoperable and Reusable.

AstraZeneca’s R&D and IT groups are partnering closely to create an industry-leading enterprise data and AI architecture. This will help us answer key business questions and enhance our ability to leverage tools, such as AI and machine learning, both now and in the future.

To this end, we are mobilising a team of data scientists, bioinformaticians, data engineers and machine learning experts from across the company to ensure we are collecting, organising and using the right data, in the best way.



Pushing the boundaries of science through AI expertise


Our leading scientists are using AI to help redefine medical science in the quest for new and better ways to discover, test and accelerate the potential medicines of tomorrow.






Collaborating to help answer big questions in AI


We know the best science doesn’t happen in isolation which is why we work collaboratively and open doors to fuel scientific discovery.



Our collaboration with Schrödinger uses their advanced computing platform with the aim of accelerating drug discovery. By combining physics-based modelling and machine learning, we will be able to predict the affinity of large libraries of potential drug molecules to identify the highest affinity candidates for synthesis and biological testing.


We have joined the Machine Learning in Pharmaceutical Discovery and Synthesis (MLPDS) consortium, an academic/industry consortium with MIT and a number of other pharmaceutical companies. The goal of the consortium is to leverage the respective expertise of the consortium members to design and deliver software tools that predict molecular properties and synthetic routes to increase the speed and efficiency of drug discovery.



We are part of a new consortium of pharmaceutical, technology and academic partners called “MELLODDY” (Machine Learning Ledger Orchestration for Drug Discovery). The project aims to leverage the world’s largest collection of small molecules with known biochemical or cellular activity to enable more accurate predictive models and increase efficiencies in drug discovery.




We are collaborating with BenevolentAI to use machine learning and AI to discover potential new drugs for chronic kidney disease and idiopathic pulmonary fibrosis. By combining our disease area expertise and large, diverse datasets with BenevolentAI’s leading AI and machine learning capabilities we hope to improve our understanding of complex disease biology and more quickly identify new potential drug targets.


AI Innovation of Sweden brings together industry, academia and the public sector in a unique partnership to accelerate applied AI research and innovation through collaboration and cross-industry sharing.