Bridging the Gap: Correlative Imaging Across Length Scales from Molecules to Organisms - From Bench to Bedside and Back Again

Cancer is one of the main causes of mortality, disease, and disability worldwide. While targeted therapy with monoclonal antibodies and small-molecule drugs has changed the natural history of some types of cancer, there are still vast areas of unmet need, where the mechanisms that allow cancer to start, progress and resist therapy are still unknown or are only beginning to be discovered.
Hospitals worldwide hold in their archives and biobanks a treasure trove of data about cancer patients. These "cancer motherlodes" have only been mined very superficially so far, as the data that makes them up is very complex, and a lot of factors need to be examined at the same time. The same challenge applies to the biobanks: generally only a very small part of the sample taken from each patient is examined because the process is very labour-intensive and requires the input of experts.
Machine Learning, the process by which computers can be made to classify large amounts of data really quickly using training sets of pre-analysed data as templates, provides a unique opportunity to finally mine cancer databases to the fullest extent.
However, by itself this is not enough: the data added to the databases needs to be of the highest quality and has to be produced fast from enormous amounts of samples.
To do so, we will develop new imaging techniques which will allow us to image biobank samples quickly and deeply, using innovative methods to look at fluorescent molecules which can tell doctors about the growth rate of cancer, or the penetration of drugs inside a tumour. We will also develop software that will allow us to "do more with less" in radiology, producing better quality images from lower doses of X-rays. The mining of these datasets will tell us which proteins and chains of reactions are responsible for making cancer patients sick and preventing them from getting better with therapy. However, knowing about which proteins are responsible is not enough. We need to know how these proteins work at different scales to make cells go rogue and become tumours. We will do this by developing a set of fluorescence and electron microscopy techniques which will work together to allow us to look at tumour proteins at increasingly larger magnification and from different perspectives, so we can look at their structure and their function together, understand how they work and figure out how to make them stop working. To do this, we will need to develop new instruments that work at very cold temperatures and in vacuum, the software needed to run them, and mechanisms to handle very cold samples without damage to them - or to us! All of this is very delicate work, and at the moment it is very slow and tricky, but the cancer patients cannot wait, so we will work together with Machine Learning experts to make taking the data and moving the samples and data between instruments quick, reliable and automatic. This way we will be able to go through more samples in a shorter amount of time and we will be able to tell the doctors in our team how the proteins they have found work and how to put a spanner in their works.
By working together, we will be able to attack the problem from multiple fronts and make headway faster.