Skip to main content
Phase III+: The University is open for expanded research operations; only authorized personnel will be admitted on campus. More info here.


January 4, 2022: (Special Time: 10AM via Zoom Only, Physics Fellow Candidate) – David Hathcock, Cornell University
Universal Absorption-Time Distributions In Evolutionary Dynamics And Epidemics.
Host: Bertrand Ottino-Loffler
Stochastic models in evolutionary biology and epidemiology often consist of birth-death dynamics where absorption times are the key quantity of interest: how long does it take for an advantageous mutation to become fixed or for an epidemic to subside? In this talk I will discuss our recent efforts to classify universal absorption-time distributions for birth-death Markov chains with an absorbing boundary state. Based on generic features of the transition rates, the asymptotic distribution for “extinction-prone” chains is either Gaussian, Gumbel, or a convolution of Gumbel distributions. In particular, the distribution is Gaussian if the transition rates are sufficiently uniform. Conversely, the later cases are closely related to extreme value theory: the Gumbel distribution emerges due to extremal events dominating the absorption process. Our classification applies to simple birth-death models of evolution, ecology, and epidemiology, but also captures the essential features of more complicated systems. For example, the distribution of times to eradicate African sleeping sickness, recently predicted using a high-dimensional epidemiological model, closely resembles the Gumbel distribution.
January 6, 2022: (Special Time: Noon via Zoom Only, Physics Fellow Candidate) – Nicholas Lammers, University of California, Berkeley
Uncovering The Kinetic Fingerprints Of Transcriptional Control Using Gene Expression Dynamics.
Host: Jasmine Nirody
Gene regulation is central to cellular function. Yet, despite decades of biochemical and genetic studies that have established a reasonably complete “parts list” of the molecular components required for eukaryotic transcription, we nonetheless lack quantitative models that can predict how these pieces interact in space and time to give rise to robust gene regulatory logic. For this talk, I will survey three distinct, yet interrelated projects from my Ph.D. that combine live imaging, computational methods, and theoretical modeling to dissect the molecular underpinnings of transcriptional control in the developing fruit fly embryo. To begin, I discuss results from a project that utilizes a novel statistical technique and live single-cell measurements of transcription to uncover how transcription factors modulate the kinetics of the transcriptional cycle to produce a sharp stripe of gene expression. Next, I share recent results that utilize cutting-edge optogenetic methods to rapidly export repressor proteins from cells, revealing that transcriptional repression—and the development trajectories it dictates—is rapidly reversible. To close, I outline ongoing theoretical work that moves beyond phenomenological models of transcription to consider a molecular picture of how transcription factor binding transmits information to drive cellular decisions. These calculations reveal that non-equilibrium gene regulatory mechanisms, which require the expenditure of biochemical energy, may be necessary in order for gene loci to function in the context of crowded cellular environments.
January 13, 2022: (Special Time: Noon via Zoom Only, Physics Fellow Candidate) – Guruprasad Raghavan, California Institute of Technology
Engineering Flexible Machine Learning Systems Inspired By Biological Intelligence.
Host: Liat Shenhav
AI systems have achieved human performance on tasks ranging from image recognition to game playing. However, they lack flexibility, a key component of intelligence. In this talk, I will focus on two facets of flexibility lacking in deep networks and will present insights from biological intelligence and differential geometry to build next-gen AI. Deep networks require laborious re-engineering, as they rely on human-designed architectures, for processing static, dynamic visual inputs from a wide range of sensor geometries. Inspired by the development of neural circuits, I present a bio-inspired algorithm that uses spontaneous spatiotemporal activity waves to grow and self-organize spiking networks on arbitrary 2D/3D sensor geometries and can internalize dynamic input representations. Additionally, deep networks cannot flexibly modify their architecture while maintaining task performance, succumbing to catastrophic forgetting (CF) while learning sequential tasks. However, humans and animals have flexible neural circuits that aren’t subject to CF. To understand principles underlying architectural flexibility, I present a geometric framework that discovers sets of networks with near equivalent functional performance on a specific task, enabling networks to preserve previous information while learning new tasks. Finally, I will present some recent work extending these frameworks to grow systems with multiple brain-regions to simulate a ‘Theory of Mind’ and enable cognitive flexibility .
January 18, 2022: (Special Time: 10AM via Zoom Only, Physics Fellow Candidate) – Sean Fancher, University of Pennsylvania
Stochastic Network Theory: Precision, Robustness, and Information Flow within Developing and Dynamic Systems.
Host: Ben Weiner
Transmission of material and information between cells is a key component of any multicellular organism. This broad class of processes manifests in a multitude of different forms at various stages of the organism’s life. In this talk, I will explore two particular examples of such material and information transmission: the distribution of morphogen molecules within developing systems and the dynamics of blood flow through the animal vasculature. In the case of the former, morphogens are particular molecular species that cells use to decode their position within the organism and thus what cell types to differentiate into. By examining two distinct models of how these molecules are distributed from their source, I will show that there exists a significant difference in the precision with which these models overcome the inherently stochastic nature of cellular processes. In the case of the latter, the blood vasculature forms a complex network of vessels that are each compliant and capable of deforming in response to changes in blood pressure. I will show how this compliance affects the speed at which information can be mechanically transmitted throughout the network by examining network topology and extrapolating analytically calculable properties of single vessels. Finally, I will discuss the basis of Stochastic Network Theory, wherein I apply stochastic dynamics to network structures, and explore its potential applications to the study of the cytoneme network model of morphogen transport.
February 15, 2022: (4PM) – Francis Corson, Ecole Normale Superieure
Mechanics Of Embryonic Self-Organization.
Host: Eric Siggia
Mechanical forces play an essential role in development, most evidently as the drivers of morphogenesis, but also potentially as long-range signals contributing to embryonic self-organization. Regulative development is particularly evident in experiments in which the avian embryonic disk is bisected: each half can give rise to a fully-formed embryo, implying a dramatic redirection of force generation and gene expression. Having identified a contractile ring, at the boundary between the embryonic and extraembryonic territories, as the engine of early avian morphogenesis, we wondered whether tension along the embryo margin might underlie embryonic regulation. Indeed, a mechanical analog of a Turing model, in which contractility plays the role of activator, and tension the role of inhibitor, recapitulates the steady pattern of tissue motion in intact embryos and its redirection upon bisection. We further show that mechanical feedback also impinges on gene expression, driving the emergence of ectopic embryos and the accompanying rescaling of embryonic territories. Our findings demonstrate a central role for mechanical forces in embryonic self-organization and cell fate allocation.
March 8, 2022: (4PM) – Stanislav Y. Shvartsman, Princeton University
Developmental Effects Of Mutations In Signaling Systems.
Host: Eric Siggia
Gain-of-function mutations affecting the highly conserved Ras cascade are associated with a range of developmental abnormalities, including heart defects, stunted growth, and neurocognitive deficits. Sequencing of affected individuals identified multiple variants in well-studied proteins, but their causal connections to the emerging phenotypes remain to be established. This is critical for our fundamental understanding of developmental diseases and requires robust approaches for following large cohorts of developing organisms with carefully controlled genotypes. We used CRISPR/Cas9 gene editing to realize such an approach in Drosophila, which offers numerous advantages for investigating organismal effects of deregulated signaling. Our analysis of developmental progression in gene-edited flies revealed that mutations compatible with development can nevertheless drastically reduce the probability of reaching adulthood in affected individuals. This quantitative result sheds light on the phenotypic heterogeneity of developmental defects caused by mutations in signaling systems.
March 29, 2022: (4PM) – Gregoire Altan-Bonnet, NCI, NIH
Universal Antigen Encoding Of T Cell Activation From High Dimensional Cytokine Dynamics.
Host: Eric Siggia
Systems Immunology lacks a framework to derive theoretical understanding from high-dimensional datasets. We combined a robotic platform with machine learning to experimentally measure and theoretically model CD8+ T cell activation. High-dimensional cytokine dynamics could be compressed onto a low-dimensional latent space in an antigen-specific manner (so-called “antigen encoding”). We used antigen encoding to model and reconstruct patterns of T cell immune activation. The model delineated 6 classes of antigens eliciting distinct T cell responses. We generalized antigen encoding to multiple immune settings, including drug perturbations and activation of chimeric antigen receptor T cells. Such universal antigen encoding for T cell activation may enable further modeling of immune responses and their rational manipulation to optimize immunotherapies. [Collaboration with Paul François’s group @ McGill – Manuscript (Science, in press): available on demand]
April 5, 2022: (4PM) – Ben Machta, Yale University
Criticality And Dynamical Bifurcations For Signal Processing And Amplification .
Host: Ben Weiner

Critical points are special places in parameter space where macroscopic system behavior is particularly sensitive to small changes in molecular details. In this talk I will explore two mechanisms through which cells could use a critical point and a dynamical bifurcation, respectively, to sensitively amplify and integrate information arriving at many individually noisy receptors.

First I will discuss the pit organ of certain snakes which forms a low resolution infrared image used for hunting prey. For this organ to be functional, individual neurons innervating the pit must be able to respond to local heating <1mk, at least 1000x more sensitive than the thermo-TRP ion channels which act as molecular receptors. I will present a model for how this remarkable sensitivity can be achieved. In this model, individual channels are electrically coupled, self-tuning to a bifurcation separating a mostly silent ‘irregular spiking’ state from an active ‘regular spiking’ state. This coupling effectively integrates an order one fraction of the information available in individual channels into the cooperative output of spike frequency.

Second I will discuss our recent work to understand the role of liquid-liquid criticality in signal transduction in eukaryotic cells. This work is motivated by two complementary experimental findings – first, that cellular membranes are two dimensional liquids tuned near a miscibility critical point, and second that many signal cascades begin with the formation of signaling cluster – a membrane domain enriched in particular lipids and membrane bound proteins under which bulk proteins with roles in signaling aggregate. I will argue that these platforms are examples of what we term surface densities; a two-dimensional prewet phase held together by critical casimir forces in the membrane and weak protein-protein interactions that reflect the propensity of many bulk proteins to phase separate. The liquid environment in these surface phases enables certain chemical reactions that are important in the early stages of signaling. As a result, the formation of a prewet phase acts as a crucial step for integrating a signal.

April 19, 2022: (4PM – Via Zoom Only) – Bjorn Sandstede, Brown University
Agent-Based Modeling And Topological Data Analysis Of Zebrafish Patterns.
Host: Bertrand Ottino-Loffler
Patterns are widespread in nature and often form during early development due to the self-organization of cells or other independent agents. One example are zebrafish (Danio rerio): wild-type zebrafish have regular black and gold stripes, while mutants and other fish feature spotty and patchy patterns. Qualitatively, these patterns display impressive consistency and redundancy, yet variability inevitably exists on both microscopic and macroscopic scales. I will first discuss an agent-based model that suggests that both consistency and richness of patterning on zebrafish stems from the presence of redundancy in iridophore interactions. In the second part of my talk, I will focus on how we can quantify features and variability of patterns to facilitate predictive analyses. I will discuss an approach based on topological data analysis for quantifying both agent-level features and global pattern attributes on a large scale. The proposed methodology is able to quantify the differential impact of stochasticity in cell interactions on wild-type and mutant patterns and predicts stripe and spot statistics as a function of varying cellular communication. This is joint work with Alexandria Volkening and Melissa McGuirl.
April 26, 2022: (4PM) – Steven Strogatz, Cornell University
The Math And Science Of Getting In Sync.
Host: Bertrand Ottino-Loffler
Every night along the tidal rivers of Malaysia, thousands of male fireflies congregate in the mangrove trees and flash on and off in unison. Similar feats of synchronization occur throughout the natural world and in our own bodies, as well as in many kinds of physical systems, ranging from pendulum clocks and metronomes to lasers and Josephson junctions. In the first part of this talk, I’ll provide an introduction to the math and science of collective synchronization. After that I’ll discuss some exciting new results and unsolved problems about how the topology of a network affects its tendency to synchronize.
May 3, 2022: (4PM) – Oded Rechavi, Tel Aviv University
Rethinking Inheritance (Of Acquired Traits).
Host: Liat Shenhav
In C. elegans nematodes, dedicated machinery enables transmission of small RNAs which regulate gene expression across multiple generations, independently of changes to the DNA sequence. Different environmental challenges, including exposure to starvation, pathogens, and heat stress generate heritable small RNA responses, that in certain cases can be adaptive. Recently we have also shown that even neuronal activity can produce small RNA-mediated heritable responses, and that the decisions that the progeny makes are affected by whether their ancestors experienced stress or not. I will discuss the underlying mechanisms, and the potential of small RNA inheritance to affect the worms’ fate. Lastly, we will examine how these new findings might affect our view of the process of evolution and the limits of inheritance.
May 10, 2022: (4PM) – Paul Francois, McGill University
Clock, Wave, Entrainment For Vertebrate Segmentation.
Host: Eric Siggia
Vertebrae precursors (called somites) sequentially form during embryonic development under the control of coupled cellular oscillators. Waves of genetic expressions (associated to Notch and Wnt signaling pathways) sweep the growing embryo from posterior to anterior, controlling somite patterning and their segregation. This complex process is highly conserved at the phenotypic level, but different genes oscillate in different species, which presents a major challenge for its understanding. I will describe how we built higher level, phenotypic models to describe vertebrate segmentation. Advances in experimental techniques further allow us to monitor and control those oscillations, leading to insights into the geometric properties of the oscillators implicated. In particular I will describe how the “segmentation clock” can be entrained and how tail bud cells synchronize.
May 24, 2022: (4PM) – Jean-Pierre Eckmann, University of Geneva
The Predictive Power Of Theoretical Biology: An Example.
Host: Eric Siggia
I will argue, supported by an experiment, that theoretical biology moves (very slowly) in a direction where predictions (as in theoretical physics) become possible. Starting from insights for a simple mechanical-genetic model of protein and the interpretation of its spectral properties, one can formulate some predictions on the presence or absence of an effect when the protein is mutated. The results suggest that only mutations at specific positions in the gene sequence have a critical effect on the function. These predictions have then been checked in a delicate experiment with actual mutations in the protein Guanylate kinase. A clear signal confirms the theoretical prediction. Some of the people involved in this work are Tsvi Tlusty, Marta Siek, John McBride, Jacques Rougemont, Elisha Moses, Eyal Weinreb, Albert Libchaber, and Stan Leibler.
September 13, 2022: (4PM) – Massimo Vergassola, Ecole Normale Superieure
Biomimetic Navigation of Complex Natural Environments.
Host: Eric Siggia
To Come.


Past Seminars

Click here for past seminars from the Center for Studies in Physics and Biology.

Contact Us

Center for Studies in Physics and Biology
The Rockefeller University
1230 York Avenue
New York, NY 10065

$('.class-name').on('click', (evt) => { $(this).hide(); // does not run a DOM query $('.class-name').hide() // runs a DOM query });