Cataloging a year of blogging: cancer and biology
January 6, 2015 15 Comments
Welcome to 111101111.
Another year has come to an end, and it is time to embrace tradition and reflect on the past twelve months. In fact, I will try to do one better and start a new tradition: cataloging a year of blogging.
Last year, I split up the 83 content heavy posts of 2013 into nine categories in three themes: established applications of evolutionary game theory (ethnocentrism and the public good; and mathematical oncology), expanding from behavior to society and mind (representations and rationality for replicators; feedback between finance & economics and ecology & evolution; and, learning, intelligence, and the social brain), and envisioning the algorithmic world (proof, automata, and physics; natural algorithms and biology; fitness landscapes and evolutionary equilibria; and, metamodeling and the (algorithmic) philosophy of science). In 2014 there was a sharp decrease in number of posts with only 44 articles of new content (and the 3 posts cataloging 2013, so 47 total) — this was due to a nearly 4 month blogging silence in the middle of the year — but a quarter increase in readership with 151,493 views compared to 2013’s 119,935 views. This time, I will need only two posts to survey the past year; this post for the practical and the next for the philosophical.
For me, the year was distributed between three cities, the usual suspects of Montreal and New York, and in October I moved down to Tampa, Florida to work with David Basanta and Jacob Scott in the Intergrated Mathematical Oncology department of the H. Lee Moffitt Cancer Center and Research Institute. A winter without snow is strange but wearing shorts in December makes up for it; plus the sunsets over the Gulf of Mexico are absolutely beautiful. Unsurprisingly, this move has meant that the practical aspects of my focus have shifted almost completely to biology; cancer, in particular.
This post is about the biology and oncology articles that made up about half of last year’s content. Given the autobiographical turn of this post, it will be (loosely) structured around three workshops that I attended in 2014, and the online conversations and collaborations that TheEGG was a host to.
Computational theories of evolution
I believe — and repeat like a broken record player — that the methods of theoretical computer science are a neglected tool (compared to the tools of physics; statistical mechanics, in particular) that has a lot to offer biology. In particular, I think that cstheory is essential for understanding learning, evolution, and their interaction. One of the most interesting topics for me in this area is the Baldwin effect (Baldwin, 1886; Simpson, 1953), which fits under the broader framework of phenotypic plasticity:
- Phenotypic plasticity, learning, and evolution (February 4th, 2014)
- Misleading models: “How learning can guide evolution” (February 7th, 2014)
- Evolution is a special kind of (machine) learning (February 14th, 2014)
- Computational theories of evolution (March 16th, 2014)
- Algorithmic Darwinism (March 25th, 2014)
Phenotypic plasticity is not only for the aspiration of applying cstheory to biology, but also one of the main to-improve-on points for the modern evolutionary synthesis. It is also of great interest for studying cancer, although I have not yet written much on that connection. Here at Moffitt, Dan Nichol is doing a lot of thinking on phenotypic plasticity — especially bet hedging — and at times taking the algorithmic perspective by thinking about how and what cell cycle switches compute (Cardelli & Csikász-Nagy, 2012). Hopefully I will convince him to contribute some guest posts to TheEGG in 2015.
Given that phenotypic plasticity and learning are sore points for the modern synthesis, it has been studied extensively. However, some of the old and celebrated models (e.g. Hinton & Nowlan, 1987) do not hold up well when peered at through the algorithmic lens. In fact, I think they might be completely misleading; attributing a ‘speed-up’ to an error in counting on the part of the modelers.
To avoid such errors, it is important to have a good formal model in which to look at evolutionary dynamics. Since there is interest in incorporating (or comparing to) learning then why not adapt an existing model from computational learning theory? This is precisely what Leslie Valiant (2009) did and thus gave cstheorists a formal invitation to evolutionary biology.
This invitation was taken up by cstheorists, physicists, and biologists at the Simmons Institute’s workshop on computational theories of evolution. An event I had the please of attending between March 17th and 21st in Berkeley; my first time visiting California. There was a wide range of topics covered and perspectives offered, and although my conference notes are extensive, I have only posted this introductory article and one follow up:
Leslie Valiant’s algorithmic approach to evolution was a major topic at the conference, with 5 different presenters discussing the model and their results or extensions. However, I thought that Valiant’s name of ‘evolvability’ conflicted with the more common use of the term in biology; a conflict that lead to some confusion during the workshop, given that several of the other attendees spoke on evolvability in the more traditional sense. In the above article, I suggested ‘algorithmic Darwinism’ as an alternative and advocated — along with Chrisantha Fernando during the workshop — for using the model as a lower bound for what is evolution can achieve without the phenotypic plasticity mechanisms of the sort we see in the Baldwin effect.
Toward the end of the workshop, I presented my work on the complexity of evolutionary equilibria (Kaznatcheev, 2013) along with some approximability results that I have not published yet:
I’ll be updating my evolutionary equilibrium work on in the following months, so you should expect more posts on this topic in 2015. But if you can’t wait then want to watch all the great talks from the workshop; there is a YouTube playlist of them. Let me know if you want me to write more about a specific topic that was discussed.
These 5 posts had a corpus of around 7.8 thousand words and garnered around 10.0 thousand views
Ecology and evolution of cancer
- Approximating spatial structure with the Ohtsuki-Nowak transform (February 26th, 2014)
- From heuristics to abductions in mathematical oncology (March 12th, 2014)
- Colon cancer, mathematical time travel, and questioning the sequential mutation model (September 16th, 2014)
- Experimental and comparative oncology: zebrafish, dogs, elephants (September 18th, 2014)
- Ecology of cancer: mimicry, eco-engineers, morphostats, and nutrition (October 9th, 2014)
- Stem cells, branching processes and stochasticity in cancer (October 25th, 2014)
My talk at the workshop was structured in two parts. First was my classification of four types of models, of which three are from 2013 and although the fourth category of abductions was suggested right away by Ishanu Chattopadhyay, I didn’t fully incorporate it into my thinking until March of 2014. This part was the more popular one. The second part concentrated on my work with David Basanta and Jacob Scott on edge effects in solid tumours (Kaznatcheev et al., 2014). Although most of the work on this project was done in 2013, and my preliminary explorations of the Ohtsuki-Nowak transform date to 2012, there was still more thinking to be done this past year.
Much like with the 2013 workshop on natural algorithms and the sciences, my goal was to produce a list of articles summarizing all the talks. Unfortunately, my will power is much higher in prime-number years, so the project is incomplete but this means that you have more posts on the ecology and evolution of cancer to look forward to in 2015.
These 6 posts had a corpus of around 9.0 thousand words and garnered around 2.5 thousand views
Viruses in cancer
- Dogs are hosts to the oldest and most widely disseminated cancer (January 24th, 2014)
- From H. pylori to Spanish colonialism: the scales of cancer (November 18th, 2014)
- Helicobacter pylori and stem cells in the gastric crypt (November 24th, 2014)
- Diversity working together: cancer, immune system, and microbiome by Jill Gallaher (December 12th, 2014)
Before moving down to Tampa this past October, I had visited Moffitt twice. During the first visit, we put together our basic results on edge effects. The second was for an annual tradition at the integrated mathematical oncology department: a hackaton/workshop in which four teams compete to model four different (but thematically related) cancer related topics. With the top team being awarded a $50,000 pilot grant to continue their research. In 2013, I was on David Basanta’s team thinking about chronic myeloid leukemia:
This past year, between November 17th and 21st, I was on Heiko Enderling’s team and the theme was viruses, microbes, and other transmittable vectors in cancer. This was a topic I knew little about, although serendipitously, I had written about the vaguely related concept of transmissible cancers earlier in the year:
As always, it was a great learning experience and we accomplished a lot in a few days. Our team secured first place, and I am looking forward to seeing where the project goes. If the above posts get you excited then you should consider joining the Moffitt team by applying to be Heiko’s postdoc (more info on his site).
This year, Jill Gallaher — the leader for team microbiome — also contributed a post giving another perspective on the workshop. I am curious to know where they’ll go from here.
These 4 posts had a corpus of around 6.6 thousand words and garnered around 1.9 thousand views
Collaboration outside the workshop
- Cooperation, enzymes, and the origin of life by Eric Bolo (February 27th, 2014)
- Misleading models in mathematical oncology (March 5th, 2014)
- Bernstein polynomials and non-linear public goods in tumours (November 7th, 2014)
- Is cancer really a game? by Philip Gerlee and Philipp Altrock (December 1st, 2014)
- Memes, compound strategies, and factoring the replicator equation (December 3rd, 2014)
Workshops are meant to inspire collaboration and seed ideas for future and current projects. But obviously, not all collaboration is from workshops. ThEGG is intended to be a place where we can bring ideas together, and so I’ve tried to encourage discussions and guest contributions here:
Although Eric’s post is not about cancer, I think it is a great point for starting to talk about cellularity and then multi-cellularity. The latter concept is a natural dual to cancer, and thus important to understand. Hopefully I will be able to convince Eric to pursue this thread further in 2015.
Early in the year, Philip Gerlee asked if there is a single landmark discovery that justifies mathematical oncology, Heiko Enderling suggested Michor et al. (2005) as a candidate, and I had to disagree. I definitely think mathematical oncology is very important and insightful, but I am not sure if Philip, Heiko, and I are using the same metrics for those qualities.
As part of a project that Robert Vander Veldge, David Basanta, Jacob Scott, and I are starting, I reviewed some recent papers on public goods games in cancer (Archetti, 2013; 2014). This prompted an active discussion on twitter and then a joint commentary from Philip Gerlee and Philipp Altrock on the (lack of) usefulness of EGT models in mathematical oncology. This post has one of the most exciting discussion threads of the year, and I recommend reading through it. Look for my response in blog post form in the coming weeks.
Although the post questioning EGT was stimulating for discussion, it did not decrease my interest in cute EGT tricks. So I closed off my posting in 2014 by sharing a trick that I’ve found useful in our project on public goods in cancer. The article itself is not related to cancer, and has a spin that relates it back to this posts’ opening topic of evolution and learning.
These 5 posts had a corpus of around 8.5 thousand words and garnered around 2.7 thousand views
Archetti, M. (2013). Evolutionary game theory of growth factor production: implications for tumour heterogeneity and resistance to therapies. British Journal of Cancer, 109(4): 1056-1062.
Archetti, M. (2014). Evolutionary dynamics of the Warburg effect: glycolysis as a collective action problem among cancer cells. Journal of Theoretical Biology, 341: 1-8.
Baldwin, J.M. (1886). A new factor in evolution. Amer. Nat., 30: 441-451, 536-553.
Cardelli, L., & Csikász-Nagy, A. (2012). The cell cycle switch computes approximate majority. Scientific Reports, 2.
Hinton, G.E., & Nowlan, S.J. (1987). How learning can guide evolution. Complex Systems, 1(3), 495-502
Kaznatcheev, A. (2013). Complexity of evolutionary equilibria in static fitness landscapes. ArXiv: 1308.5094v1
Kaznatcheev, A., Scott, J.G., & Basanta, D. (2014). Edge effects in game theoretic dynamics of spatially structured tumours. arXiv arXiv: 1307.6914v2
Michor, F., Hughes, T., Iwasa, Y., Branford, S., Shah, N., Sawyers, C., & Nowak, M.A. (2005). Dynamics of chronic myeloid leukaemia. Nature, 435(7046): 1267-1270.
Simpson, G.G. (1953). The Baldwin effect. Evolution, 7(2): 110-117.
Valiant, L.G. (2009). Evolvability. Journal of the ACM, 56(1).