by ecancer reporter Clare Sansom

One of the over-arching themes at the 2015 National Cancer Research Institute conference in Liverpool is the importance of 'big data' for cancer research and treatment.

This was picked up in different ways on the first morning of the conference by both keynote speakers and in a well attended symposium.

The first plenary speaker was Fabrice André from the Institut Gustave Roussy, Paris, France, who addressed the problem of selecting and proving the efficacy of personalised cancer therapies.

He started by giving four reasons why so many clinical trials of drugs tailored to specific mutations have failed to prove efficacy.

These were, briefly, the fact that many of the trials were designed to discover new combinations of mutation and drug rather than show efficacy; the difficulty of matching a drug with a genomic alteration; the frequency of driver mutations; and difficulties with single-agent trials.

Clinical trials of personalised therapies can be divided into three types based on the level of validated evidence for the target and the frequency of the mutation concerned.

Identifying the subset of patients predicted to have a very poor outcome with any current therapy presents a particular challenge.

We don't yet have all the tools and methodologies we need to identify the important 'driver' mutations in tumours, and we need much comprehensive clinical and genomic datasets in order to test these.

The second plenary speaker, Amy Aberbethy from Flatiron Health, a Google company based in New York, USA, gave a compelling presentation of the importance of putting the cancer patient at the centre of big data initiatives.

Her talk was centred around the experiences of an individual melanoma patient she called Janet: a woman in her 30s with fair skin, freckles, and a family history of the disease.

Janet would typically have only a 3% chance of being involved in a clinical trial, and even that would fail to capture all the details that could be of value for research.

A patient's complete story from presentation to cure (or death) will produce an enormous quantity of disparate data that could all be useful in research and clinical practice.

“Data is a precious resource, and unlike other such resources – gold, for example – it doesn't get used up; instead it can become more valuable over time”, said Abernethy.

She explained the need for collecting and storing data in a standard format so that, for example, patient outcomes and genomic data can be combined and and incorporated in an electronic patient record, and made an eloquent case for involving patients in all aspects of this process.

Although this may be too late to help Janet and her fellow patients in today's oncology clinics, it should prove invaluable to future generations.

The talk was followed by a lively discussion that was brought down to earth by an impassioned plea from one of the patient representatives at the meeting: the wife of a myeloma patient who explained that her husband's records were not yet shared even between neighbouring hospitals in Wales.

This contribution, which received warm applause, highlighted the difference between 'best practice' in dealing with cancer patients' data and what they experience too often.

Tumour genomics was also a main theme in a symposium entitled 'Big Data', which was chaired by Nicholas Luscombe from Cancer Research UK's London Research Institute (shortly moving to the new Crick Institute in London).

Paul Boutros from the Ontario Institute for Cancer Research, Toronto, Canada described how  genome sequencing of individual prostate tumours is helping elucidate the complex and heterogeneous nature of these tumours.

Prostate cancer is extremely common and is now very easy to diagnose, but it is much harder to distinguish between indolent tumours that can be treated by 'watchful waiting' and more aggressive ones.

Boutros showed that many prostate tumours are so genetically heterogeneous that some men appear to have developed two (or perhaps more) distinct cancers.

Copy number variation seems to be the most promising biomarker for predicting prognosis, and combining analyses from many independent groups can often give a more accurate prognosis than any single one.

Peter van Loo, also from the Crick Institute, described genetic changes in tumours in terms of what he called their 'molecular archaeology'.

Mutations accumulate over time in all cells, and tumours arise when a cell population accumulates enough mutations to drive an uncontrolled expansion.

When a tumour is characterised it can contain many 'sub-clonal' populations of cells with the same mutations, and it is possible to identify a 'last common ancestor' with only those mutations found in the whole of the tumour.

The International Cancer Genome Consortium is coordinating projects to describe genomic, transcriptomic and epigenetic changes in 50 different tumour types, and it has already collected over 800 Tb of data.

Genomic heterogeneity is known to vary considerably between tumour types, but melanoma is the only type in which over 50% of tumours have been found to consist of a single clone.

Tracking the evolution of clones throughout the lifetime of a tumour may reveal new treatment options.