SIAM Life Science Conference in Minneapolis

Here are some impressions from the SIAM Life Science Conference. On Monay I heard a talk by Nathan Kutz about the structure and function of neural networks in C. elegans and a moth whose name I have forgotten. Coincidentally, when I was in Boston I heard a presentation by someone in the Gunawardena group about genes and transcription factors related to different types of neurons in C. elegans. On that occasion I asked how many of the few hundred neurons of C. elegans are regarded as being of different types and I was suprised to hear that it of the order of one hundred. Returning to the talk of Kutz, the key thing is that all the neurons in C. elegans and all their connections are known. Thus it is possible to simulate the whole network and reproduce central aspects of the behaviour of the worm. Another simplifying circumstance is that the motion of the worm can be described by only four parameters. A central message of the talk was that the behaviour of the nervous system of the worm itself can be reduced to a low dimensional dynamical system. This reminded me, with how much justification I do not know, of what I heard about the beaks of Darwin’s finches in Lisbon recently. As for the moth, the question described in the talk was how it learns to identify odours. A lot is known about the architecture of the neurons in the moth’s olfactory system. The speaker has compared the learning ability of this biological neural network with that of artifical neural networks used in machine learning. In reaching quite a high accuracy (70%) on the basis a small amount of training data the moth network beat all the artifical networks clearly. When given more data the result of the moth network hardly improved while those of the artificial network got better and better and eventually overtook the moth. The speaker suggested that the moth system is very effectively optimized for the task it is supposed to perform. I also heard a nice talk by Richard Bertram about canards in a three-dimensional ODE system describing the electrical activity of heart cells. I found what he explained about canards enlightening and this has convinced me that it is time for me to finally understand this subject properly. In particular he explained that the oscillations associated with a canard have to do with twisting of the stable and unstable manifolds. I also found what he said about the relations between the canard and concrete biological observations (the form of electrical signals from neurons) very interesting.

On Wednesday I heard a talk by Benjamin Ribba, who after an academic past now works with Roche. One of the main themes of his talk was cancer immunotherapy but I did not feel I learned too much there. What I found more interesting was what he said in the first part of his talk about chemotherapy for low-grade glioma. This is a disease which only develops very slowly, so that even after it has been discovered the tumour may grow very slowly over a period of several years. He showed patient data for this. He explained why it was reasonable to leave patients so long without therapy. The reason is that the standard course of therapy at that time was something which could only be given to a patient once in a lifetime (presumably due to side effects). Thus it made sense to wait with the therapy until the time which was likely to increase the patients lifetime as much as possible. During the slow growth phase the disease is typically asymptomatic, so that it is not a question of years of physical suffering for the patient. The speaker develeloped a mathematical model for the evolution of the disease (a model with a few ODE) and this could be made to fit the data well. The time course of the tumour growth is as follows. Before treatment it grows steadily. After the treatment is started the tumour starts to shrink. After the treatment is stopped it continues to shrink for a long time but usually eventually takes off again after a few years. In the model it was possible to try out alternatives to the standard treatment protocol and one was found which performed a lot better. The talk did not include any indication of whether this information had been used in paractise. I asked that question at the end of the talk. The answer was that by the time the analysis had been done the standard treatment has been replaced by a very different one and so the theory was too late to have clinical consequences.

On Thursday I heard a talk by Dominik Wodarz. One of his themes was oncolytic viruses but here I want to concentrate on another one. This had to do with chronic lymphocytic leukemia. This is a B cell malignancy and is treated using ibrutinib, an inhibitor of BTK (Bruton’s tyrosine kinase). Most of the malignant B cells form a tumour in the bone marrow while a small percentage of them circulate in the blood. Measuring the latter is the easiest way of monitoring the course of the disease. When the drug is given the tumour shrinks but it was not clear whether this is because the cells leave the bone marrow for the blood, where they die, or whether a large proportion die in the marrow. Using a simple mathematical model it could be shown that the second of these is the correct explanation. There are several things I like about this work. First, very simple models are used. Second, after comparison with data, these give convincing explanations for what is happening. Third, these conclusions may actually influence clinical practise. This is a good example of what mathematical biology should be like.

Advertisement