Thinking Inside the Box: five ways the ICR is leading the way with Big Data in cancer research

Image: Big Data cloud (image: Costas Mitsopoulos , Amanda C. Schierz , Paul Workman, Bissan Al-Lazikani)

The pace of technological advancement has led to researchers biting off more data than they can chew.

Science generates so much data, containing so much hidden meaning. A major challenge for our Big Data researchers now is in thinking ‘inside the box’ – to allow us to better use the data we already have access to.

Artificial intelligence (AI) is helping us bring together data in new ways, allowing us to ask new questions and gain new understandings of cancer.

AI can learn to find hidden patterns and make informed recommendations to guide decision-making. AI is able to constantly adjust its response to the information it is fed, and the technological revolution we find ourselves in is ensuring the data are in no short supply.

Professor Andrea Sottoriva and his team are applying skills in computer science and mathematics to big datasets, seeking out the evolutionary patterns that guide cancer’s development, and will play a key role in the vibrant culture of the Centre for Cancer Drug Discovery.

Find out more

Here are five ways the ICR is helping transform the application of Big Data and AI to cancer research.

1) Drug discovery

A recent editorial from Professor Paul Workman, Dr Albert Antolin and Professor Bissan Al-Lazikani, published in Expert Opinion on Drug Discovery, explores how these technologies are impacting and transforming modern drug discovery.

In the editorial, the authors highlight how Big Data from the clinic, as well as molecular and non-clinical data, is already being applied to the identification of new drug targets, and also in the subsequent drug discovery campaigns.  

Researchers at the ICR’s new Centre for Cancer Drug Discovery will focus on finding ways to out-smart cancer’s ability to develop resistance to frontline treatments.

Big Data and AI will play a major role in identifying how cancer gets ahead, and how we can identify new drug targets which will address unmet clinical needs, with a strong focus on overcoming cancer evolution and drug resistance.

2) Repurposing drugs

Professor Bissan Al-Lazikani, our Head of Data Science, is one of our leading figures in the development of new AI tools. In her journey with Big Data and cancer at the ICR over the last 10 years, one of the highlights has been the development of canSAR, the world’s largest cancer knowledge-base.

The platform is not only able to inform researchers in drug discovery, but also to find drugs which might be able to be repurposed for a given type of cancer.

One example of this is a 2018 paper by a team of scientists from around the world, including experts from the ICR, which used canSAR to analyse genetic mutations in patients with prostate cancer. The researchers used AI to decipher the hidden circuitry of prostate cancer and recommend several existing drugs for specific types of prostate cancer.

The analysis found that there were 11 targets for which a drug already exists – in other words, they found 11 locks for which we already have a working key.

It didn’t stop there. They also found seven more ‘locks’ which have drugs that are currently being researched, and another 62 potential targets which could be considered for future research.

canSAR integrates data seamlessly and analyses it all together – from cancer cell lines, to patient tissue samples, to chemistry and 3D protein structures. It is used globally by scientists and is just one example of the ICR’s pioneering work on applying Big Data to help us understand and treat cancer, knitting together data from all across the research landscape so we can see the bigger picture.

AIs can be trained to interrogate the billions of data points in the canSAR database, and can help researchers find new cancer proteins that share hidden but important structures in existing drug targets. This means scientists can find the Achilles heels of cancers, and potentially repurpose existing drugs to exploit these newly identified weaknesses. 

The canSAR database is helping scientists all over the world in their efforts to discover new drugs and answer big questions in their research. It is currently used by nearly 100,000 cancer researchers annually.

3) Targeted radiotherapy based on patients’ needs

And that’s not all Professor Al-Lazikani’s team have been up to. Along with other researchers from several departments and Divisions across the ICR, they’re also working on improving radiotherapy treatment for patients.

Using AI on a huge amount of patient data, they are looking for patterns that could help to predict which patients are most likely to suffer long-term side effects.

Nearly two thirds of cancer patients will receive radiotherapy – a hugely effective treatment that cures many – but around 20 per cent of them will experience long-term side effects, often only developing months or years after treatment.

At the moment, there’s no way to reliably predict which patients are most at risk of developing side effects. As a result, all patients receive similar treatment regimens due to the lack of a good predictor for their individual level of risk.

Using AI, our researchers are mining vast quantities of data to try to uncover clues that could help guide radiotherapy treatment and lead to new tests.

In research presented recently at major radiotherapy research ESTRO, ICR researchers showed that by weaving together data from several different sources – including genes, radiotherapy dose maps and blood tests – they could for the first time predict which patients are at higher risk of long-term side effects.

In future, this will help doctors to better tailor radiotherapy treatment to individual patients and give them more informed options for their treatment.

4) Predicting how cancer will evolve

The most challenging aspect of treating cancers is trying to stop tumours from outsmarting the drugs we throw at them. Cancers can respond quickly to treatments and are able to evolve into drug-resistant forms – similar to the way bacterial infections are able to quickly outsmart antibiotics. 

The REVOLVER study, led by Professor Andrea Sottoriva, Deputy Director of our Centre for Evolution and Cancer at the ICR and scientists at the University of Edinburgh, has used AI to predict how cancer will change its DNA over time, predicting how cancer will evolve.

Getting a sneak peek into the future of a tumour could help scientists and clinicians act earlier, and pounce on cancer before it evolves to resist treatments.

Cancer evolution is a major focus for the ICR, and the new Centre for Cancer Drug Discovery will bring together scientists working on AI and evolution with researchers in drug discovery to address these complex research questions.

5) Using cancer images

An enormous amount of data is generated from patient imaging – a single scan can produce thousands of images. Multiply this by the thousands of patients being scanned and it’s easy to see why AI is well placed to comb through the expanse of data to look for patterns.       

Dr Yinyin Yuan, Team Leader in Computational Pathology at the ICR, is using computational models to assess images of tumour samples – a technique which in the future could pick out women with especially aggressive ovarian cancer.

Dr Yuan and her team are using an AI tool which is trained to look for clusters of cells within tumours which have a misshapen nucleus – the control centre of the cell. Women who have these cell clusters tend to have more aggressive disease than those without.

Using AI in this way is giving us access to hidden information on weaknesses in tumours, which in the future could be useful in predicting how patients will respond to treatments.

Using this information gathered through AI analyses of tumour samples, treatments might then be changed according to patients’ needs. 

Exploring inside the box

There is a long way to go – in order to be able to get the most out of data, we need common guidelines around the quality of data entered into the public domain.

Harnessing the data available is a huge challenge given its scale and complexity, but it will form an integral part of understanding and treating cancer in the future.

Drug discovery will rely on properly combining public data with an organisation’s internal data in a seamless fashion, and the development of powerful AI tools which can weave together these data to tackle the big questions in drug discovery.

We are in the throes of a revolution that is well on its way to helping identify, treat and defeat disease, and the ICR is focused on making the discoveries that defeat cancer.