Below are some thoughts on my experience at APS March meeting. Ended up missing about a day of it due to my hurt ankle, but hopefully I’ll cover most of the relevant things that happened.

Eco-evo theory

Most of the eco-evo theory work was on consumer resource models. There were a couple of talks on spatial structure stuff that I missed, but they were mostly involving pushed vs. pulled waves with one type (modeling density dependent effects). As a side note, a couple of the talks used (explicitly or just in a cartoon sense) the idea that interactions in the consumer resource model can be thought of as an outer product of matrices involving how resources affect types and how types affect resources.

Cooperation and symbiosis

There were a couple of talks trying to understand how cooperation/symbiosis/cross-feeding could be evolutionarily stable. Unfortunately most of them were very mathematical - trying to construct complicated game-theoretic models and doing even more complicated analysis.

There were two intriguing talks: the first was by Simone Pigolotti talking about bet-hedging strategies. He claimed that his modeling suggests that spatial structure is necessary for bet-hedging to evolve, might be worth taking a look. The other one was trying to look at cross-feeding; they had a simple toy model in which one resource could be broken down into a secondary resource, which could again be broken down for energy. They looked at types which could breakdown both as well as types that could only break down one, and tried to study their coexistence. The claim was that a community of two cross-feeders was stable to invasion from a “generalist” who broke down both.

Seems that there are a lot of common threads in a lot of these “cooperativity” models; I wonder if there are general principles that would be worth writing down.

Papers

Consumer resource models

The main eco-evo theory stuff that got talked about was consumer resource models. There was one talk on trying to understand the spectrum of the interaction matrix around the fixed point; was hard to follow details in the talk so I included references below. Ned Wingreen talked about his work on the linear tradeoffs model that has the degenerate subspace (missed the talk unfortunately), but didn’t really have many new results (main thing is that they are now trying to understand spatial structure).

Pankaj Mehta and associates had a few talks; probably worth reading their latest papers just to get a sense of what their models are. My impression is that their results are similar to Mikhail Tikhonov’s work, but would be good to figure out in detail - especially as there may be subtle, strong assumptions used.

Papers

Ecology data

Missed a few of the experimental talks due to my foot injury, but judging from abstracts/references, most were older ones that I’ve heard about/read about/talked to the authors about. One of the talks had some references to some work looking at abundances in an ecological system (fish) over long periods of time - which could be of interest as references for any ecology theory work on dynamical diversity.

Papers

Evolving generalists vs. specialists

One theme that came up in a couple of talks was the idea that in order to evolve generalists, you need to be presented with environments (either in time series or via spatial structure) that are different but not too different. This came up in Mikhail Tikhonov’s talk (on the model that he’s writing up right now), as well as Vedant Sachdeva’s. Something that I hadn’t thought too much about, but seems like a good fundamental idea.

Phenotype models

There was one talk dealing with evolution on phenotypic landscapes, in the context of the immune system. The model was similar to mine, so will be worth thinking about how the existence of this work might affect how I develop mine.

The talk, by Jiming Sheng, was trying to model the co-evolutionary dynamics of antigens and antibodies in the immune system. There was some phenotype space, where again some sort of distance measure was related to fitness. The model was meant to study co-evolutionary dynamics in the presence of broadly neutralizing antibodies. I think it had both dynamics and mutations, but not sure of their timescales. The main result was that depending on how broad the distribution of initial antibodies was, and the dimensionality of the phenotype space, one would either get the antigens “escaping” or being quickly neutralized.

For most of the talk, it seemed that the antibodies were not mutating (or mutating slowly), and most of their influence was via their intial distribution. However towards the end, there was a brief discussion of how the antibodies could start out in a non-broadly neutralizing state, and go to a broadly neutralizing state, but I need to get some more details about the model to figure it out.

In terms of what they haven’t done, I think one of the main things is that they don’t really have any analytical results. They haven’t thought about scaling with any sets of parameters in quantitative detail. It’s also unclear in what sort of dynamical regime they are working - there may have been a separation of timescales between the dynamics and mutations, but it was unclear from the talk. Also, there was not much discussion of the diversity of either population.

Wasn’t able to catch up with the speaker in person, unfortunately; plan on following up via email to get a better sense of the questions/model/results. No papers/pre-prints as of yet, but I’ll update this when I get more information.

There was also some talk that mentioned some work involving evolution in a phenotype landscape giving phase transitions. Aboslutely no detail in the talk, so will have to look at the papers to see if the modeling is interesting/relevant.

Papers

Immunology

There were a few talks on immunology; however, most were about modeling specific biophysical situations involved in adaptive immunity. On the evolutionary theory side, there was the above mentioned phenotype modeling talk. There was also a talk by Andreas Mayer (working with Bialek et. al.) to try to model the genetic diversity of human antibodies as compared to the DNA of pathogens, specifically by looking at the distribution of \(k\)-mers (usually for \(k = 9\)). The idea was to see if there was signal that the immune system could theoretically use to distinguish self from non-self. They used a simple pairwise interaction energy model which seemed to work ok.

There were differences between human and non-human; however, it was not clear how important those differences should be. For example, things like GC content different gave very strong signal, but pathogens could ostensibly modulate GC content in proteins that interact with antibodies. It is also unclear what sorts of “discrimination technology” the body has access too - ostensibly, there are types of difference that are hard to measure via antibody interactions. The basic idea is interesting though; I wonder if machine learning methods (or at least, machine learning ways of asking questions) might be relevant/important here, since what really might matter is the discriminability of a small number of proteins/parts of proteins, and the pathogens have the capability to make at least small changes to avoid detection.

Miscellaneous talks

Caught the tail end of a talk that was about trying to understand instabilities in reaction-diffusion type systems using tools from standard non-linear dynamics theory (including phase portraits and nullclines). There are two papers that might be worth looking at, especially in the context of the dynamical systems class.

Edo Kussell had some interesting work about trying to infer recombination rates in regimes where the recombination rate is similar to the mutation rate. They tried to do this by constructing a model which had coalescent times both for each pair of individuals in a sample, as well as an overall coalescent time. They then tried to measure the lengths of blocks that were identical (or near identical) between individuals, and use the distribution of those blocks in order to fit the timescales/rates. Seems like a nice choice for a journal club article.

Nicholas Noll (formerly with Boris Shraiman, now a post-doc with Richard Neher) had a nice talk about trying to measure the rates of different recombination events in bacteria. I thought the framing of his talk was really good - trying to motivate the question of what the “atom” of genomic analysis in the presence of high recombination should be - in the same way that a tree is the right structure in the low recombination, low diversity regime. Had some results as well, but I think the motivation was the star of this talk.

There were also a few talks on epistasis, which all gave the low dimensional, low order epistasis, rugged landscape picture. Hopefully my paper comes out soon because there are a lot of lessons there for people to internalize!

Overall impressions

It seems that the eco-evolutionary dynamics subfield is gaining a lot of steam (at least among physicists). Some of the high presence at the meeting was due to Jeff Gore organizing a bunch of sessions, but seems to be of broad interest. Theoretically, still feels as if there is some separation between the ecological aspects of the modeling, and the evolutionary aspects. The consumer resource models are very popular on the theory end; I think I need to do a bit more to catch up but as far as I can tell. I think trying to add in evolution and think of evolution-conditioning of things is a very open area.

Experimentally, it’s still not yet clear to me what the good systems are. Most of the experimental talks I was able to make it to seemed very limited in scope - mostly focussing on some particular phenomenology of interest. Unfortunately missed some more of the ecological ones so can’t comment too closely on this point.

Overall, I’m not a big fan of the statistics of APS - because there are so many short talks, it’s hard to get a good sense of any one thing. It’s also hard to discern which work is deep and careful, and which work is hiding serious flaws and limitations. Also, the sessions seemed too broad at times - for example, my talk followed one about the mechanics of parasitic worms.

Still, was nice to give a talk that I thought was pretty good, both on its own merits and relative to the surroundings. Feels good to have projects that I think are interesting scientifically!