The complexity of cancer biology has always amazed me. Regardless of how much you learn, discover or understand, it always feels like just the beginning, like nature has put up an impenetrable wall that the human mind could never break through.
My first research project as an undergraduate student in biology was looking at how one tiny molecule was controlled by another, and the role this played in nerve regeneration. I remember spending hours trying to understand how all the pieces of the puzzle fit together and why this molecule, among the hundreds of thousands of others, was special. At some point, however, I realized that it was impossible – it was just too complex for a human brain to remember all 20,000 gene names and all the pathways in which they are involved (although, I must say, I have met people who seem to have come close to this). I didn’t know it at the time, but it was at that point I decided to become a computational biologist, although I could barely write 3 lines of code!
As a cancer researcher in computational biology and bioinformatics, data (not pipettes!) are our bread and butter. People think of different things when you say data – “numbers”, “digits”, “storage” – but to me, it represents an undiscovered world of networks and patterns. One small change in one corner of a complex network could have a cascade effect and disrupt the entire system, transforming a normal cell to a cancer cell. Or, perhaps seemingly random changes in gene activity or in the cells that aid tumor growth are actually part of a consistent – albeit complicated – pattern. This is the common ground of all the projects I have worked on – I look for ‘patterns’, for ‘networks’. In other words, I try to define the ‘rules’ that govern a system.
In my first computational project we wanted to explore whether cancer cells no longer relied on being part of an organ and started resembling single-cell organisms, such as bacteria. To answer this question we were using data at the level of the activity of genes – that is, looking at what genes are switched on or off in cancers at any point in time. In practice this meant that we were looking for a ‘pattern’ among 20,000 genes and 10,000 samples (so 2x108 data points), where we didn’t know what the pattern was or how we could find it. In the end we uncovered an underlying architecture of biological networks, where key interactions were broken down by mutations. With this, we had added a novel layer to our understanding of the fundamental biology of cancer.
Networks of connections between genes involved in cancer, such as this one, often look like flowers. A coincidence or a hidden pattern?
My latest project is largely translational, where we are trying to identify prostate cancer patients likely to respond to immunotherapy. For this, we are using an exciting new type of data and analysis – spatial analysis of tissue microscopy images. Here too the computational aim is to find patterns, but of a different kind – that of how immune cells interact with tumor cells. Tissues, unfortunately, are messy, and my question here is less “what is the pattern” but rather “what can we measure and how can we do it?” They say that an image is worth a thousand words, but unfortunately, they do not give you p-values or data distributions that you could add to a publication.
We realized very quickly that we needed to build new tools to interpret this data. I am now integrating quantitative methods borrowed from the fields of geography and ecology to uncover and quantify hidden patterns without relying on the human eye. Up to now we have found that particular spatial patterns of immune cells are associated with either immunosuppression or tumor immune recognition, and we are interested in understanding how they link to disease progression, metastasis and response to treatment.
We are now in the age of big data in biology and medicine, and we certainly see regular use of genomics in the clinic, helping treating clinicians make decisions. However, to me, one of the major challenges is one of data integration across biological levels. How do the patterns of alternations we see at the DNA level translate to the cell level, then to the tissue, the organism, and end up being determinants of disease aggressiveness, patient response to treatment and prognosis? How do all these systems build up from the other? What are the overarching rules and patterns that govern these systems?
Can we, as mere humans, ever understand it? Or is it, as my undergrad self once thought, just too complex?
I’ll let my postdoc self answer: No. We just need the right mindset, questions and tools. We just need computational biology.
Dr Anna Trigos is a Postdoctoral Researcher in computational biology, bioinformatics and prostate cancer. Her expertise includes gene expression, genomics, gene networks and tissue microscopy image analysis. She won the Joseph Sambrook Prize for Research Excellence and the Postgraduate Research Medal in 2019.
Dr Trigos can be contacted by:
Email: [email protected]