There was recently a twitter discussion that I am now being held accountable to turn into a blog post by Jeffrey West (thanks?) and I have renewed interest based on a lunch with Dr. Philip Maini. I take pause because of some backlash that Rob Noble has received with his blog post of a very similar nature. His post was addressing something different I think and I largely agree with everything that he states (disclosure, I’ve never met Rob or worked with him). The purpose of his post, in his words, was to…”help clarify (for junior researchers like me) how scientists applying mathematical and computational methods to biological problems have organized themselves into communities.” Those communities being necessary for feedback, review processes, and determining appropriate conferences. From a broader point of view and working across the two fields that Rob has delineated, mathematical biology and computational biology, I think there is still much overlap and room for further explanation (from my perspective) of where we (or maybe just me, myself, and I) fit within the research ecosystem (see the only figure in this blog post above).
Where I’m approaching this from:
My goal in coming to Oxford, stripped to the core, was to work entirely in the “in-between” at different scales and build some cool agent based (preferably hybrid) models along the way. I started working with Dr. Alexander (Sandy) Anderson at Moffitt Cancer Center as a Bioinformatician and Computational Biologist (funny enough this was exactly the name of my masters, further supporting Rob’s point) after a brief stint with Roche Pharmaceuticals working as a bioinformatician for a later clinical trial for, if I recall, Alectinib. At Moffitt I started to dabble in ABMs and branching process models all with a genetic focus and the rest is history. I add this information only to hope I add a minuscule amount of credibility for what I say below.
During my time at Moffitt, the IMO workshop (keep eyes out for the next one), and the past two years at Oxford I’ve been fortunate to have many interactions with experimentalists, mathematicians, bioinformaticians, statisticians (where most of my masters cohort ended up), and clinicians. The first time I talk to experimentalists, clinicians, and bioinformaticians who are new to mechanistic modeling (a constituent of mathematical biology) they, understandably, have pre-conceived notions that what I’m going to pitch to them is some newfound statistical model. Most, if not all, take a while to wrap their head around the idea that there are multiple forms of mathematical modeling (statistical being one) that could be useful. I don’t think this comes as a surprise to many readers and at least my mathematical oncology peers have expressed similar feelings. Within the first year in my program at Oxford (Genomic Medicine and Statistics at the Wellcome Centre for Human Genetics) I had to create a succinct figure to explain where the hell I fit in any of this craziness to lead with for talks and meetings.
What is the research ecosystem
We start with the premise that we are striving to understand complex relationships and systems. These complex relationships are cell-cell interactions or environment and tissue or whatever you study. In the biomedical (my area) research ecosystem I envision an input and an output. We observe a phenomena, which together with its different datasets, becomes the primary input into this ecosystem. There’s a number of tools at our disposal. We may first try to build statistical models to determine what correlates and when to expect our observed phenomena (statistical modeling, arguably a clinicians best friend). Then our experimentalist comrades may be trying to force some in vitro cells to produce a protein that they think could be responsible for said phenomena to determine the who; if that works maybe a mouse model would be used to tell us how that protein is working and to dig deeper into the biological system it is present in. Due to the ultimate goal of prevention and treatment of the observed phenomena, drug discovery must fit somewhere it exists not as a what, when, who, or how, but rather as a necessary component that we all must consider.
The output of this research ecosystem is really what impacts the patients and the community. It’s the hope that what we do means that patients, and we, can spend a few more years with our parents, grandchildren, and devastatingly, children at times. We want to provide better prognostic indicators, better means to catch cancer quicker for diagnostic purposes, and we want to treat effectively while minimizing impact to patients lives and the larger organismal system (off target effects). We need to have ways to treat patients smarter and make sure we are utilizing proper dosages. But where does mathematical biology fit into any of this ecosystem?
Well, it’s capable of being integrated at all levels. Within our ecosystem all the players are already there that try to give us the who, what, when, where and how. Mathematical biology (mechanistic modeling is a part of this) is yet another tool, sometimes the only feasible tool, that can contribute to the how and what (forgive me, it’s not in the picture). This has some implications that I think are crucial. Every facet of this ecosystem needs to be translatable to another component of the ecosystem in some form.
As a recent example, for me this means that I will observe some data, probably sequencing data if I’m honest, that will be at a molecular level from biopsies I’ve collected. I will build a cell based model and then I will ask experimentalists (or just the literature) if the aspects of my model that govern behavior of cells make sense. However, my model has to be able to have an output directly comparable to the observed data (sequence information). I then compare this information or fit my data using statistical modeling techniques.
This is why if I had to point to my figure and say where I fit it would require an explanation of mathematical biology, statistical modeling, and experimental biology. However, I’ve also got a project that explores the dynamics of a potential new treatment compound. So I don’t exist in a single, isolated, part of the ecosystem and I challenge everyone to ensure that they don’t either. I think the best mathematical oncology projects and papers that I’ve had exposure to in my humblest of opinions make it really hard to tell where the experimental, statistical, and mechanistic modeling begins and ends. Why is that? Because it means that the who, what, when, where, and sometimes how is usually being answered in a complete story.
Finally, from the title of this, where does mathematical oncology fit in the research ecosystem? It can fit with every part of it. We just need to make sure that we help others see that. Which is the point of trying to explain the research ecosystem.