In trying to understand how RNA might catalyze its own replication, we must confront RNA motions on time scales ranging from picoseconds to days. For example, MD simulations reveal rapid shifts between preferred conformations. Metal ion binding can slow these fluctuations by stabilizing a particular conformation. This stabilized conformation can direct copying chemistry on a scale of minutes. The primer-template complexes required for copying chemistry can form and dissociate in seconds to days depending on the number of interacting species. I will describe our efforts to study these diverse phenomena as we attempt to build up an integrated picture of nonenzymatic RNA copying and replication during the origin of life.

New fossil discoveries, coupled with the analyses of genome structure and function during appendage development, reveal the ways in which novelties arise in evolution. Here we will investigate how tetrapod locomotion originally arose within a fish body plan. A regionalized axial skeleton, enhanced pelvic appendages, and appendages with the three major segments of tetrapod limbs are first seen in the lobe finned fish that are most closely related to tetrapods. Genomic analyses on extant fish reveal the deep conservation of regulatory architecture and gene functions involved with the patterning of both fins and limbs. Because these genomic and functional similarities underlie the development of both paired and unpaired appendages they also provide insights into the origin of appendages in vertebrates.

Coral reef fishes have developed a wide range of intriguing structural and functional novelties during a hundred million years of evolution. We will explore genetic and computational approaches to resolving the fish tree of life, and visualize the branching network of relationships among species in several diverse fish families. In addition we will explore the engineering linkage design of highly kinetic jaw mechanisms, the evolution of propulsor shape and material properties in fish swimming mechanisms, and the unusual sediment burrowing behaviors of diverse reef fish groups.

The APS Panel on Public Affairs (POPA) decided last year to look into the issues faced by the potential need to remove CO2 from our atmosphere, as suggested by the latest IPCC report on climate change. I have been part of the collaboration working on this project, and our report is currently under review. My talk will focus on the main issues we’ve identified and - unlike the report, which studiously avoids policy opinions and recommendations - I will comment on the policy issues that flow from this investigation.

From electrically responsive neuronal networks to the adaptive immune response, biological systems can learn to perform complex tasks. In this seminar, we explore physical learning, a framework inspired by computational learning theory and biological systems, where networks physically adapt to applied forces to adopt desired functions. Unlike traditional engineering approaches, physical learning is facilitated by physically realizable learning rules, requiring only local responses and no explicit information about the desired functionality. Our research shows that such local learning rules can be derived for broad classes of physical networks, and that physical learning is indeed physically realizable through laboratory experiments. By leveraging the advances of statistical learning theory in physical machines, we propose autonomous physical learning as a promising bridge between computational machine learning and biology, with the potential to enable the development of new classes of smart metamaterials that adapt in-situ to users’ needs.

Quantum algorithms are prized for their potential to outpace classical computing, but since quantum algorithms first started to be developed 25 years ago, classical algorithms closely mimicking quantum computers have been developed based on tensor networks. The analogy is so precise that tasks originally conceived for quantum computers, such as performing Fourier transforms of low-dimensional functions, can be carried out entirely on classical computers in many cases. Remarkably, these algorithms can be better than any previous classical algorithm for certain tasks. I will give an overview of algorithms based on quantum-inspired encodings of functions and argue that they are already useful for tackling diverse problems while using a single unified framework.

One of the pleasures of an experimentalist is the chance to try to understand how molecules think, giving nature the opportunity to open new questions though we may begin by trying to answer older ones. My lab’s experience is that experiments don’t always confirm theoretical expectations. I will give examples.

Understanding the dynamics of the immune system has been gaining increasing importance, as a consequence of progress in immunotherapy applied to cancer and due to the importance of vaccines re the COVID-19 pandemic. This talk will survey examples which show physics can help address some of the important conceptual and also practical issues that arise in this research area, specifically focusing on issues related to cancer. Specific topics to be discussed includes the detection of neoantigens as a way to target tumor cells, the spatiotemporal dynamics of immune cell infiltration into tumors, the ecology of the immune microenvironment and the role of evasion by cancer cells.

Due to their out-of-equilibrium nature, materials such as amorphous solids, glasses or dense suspensions exhibit a history-dependent mechanical behavior. Thermal and mechanical annealing drastically affect the modes of deformation and failure. In recent years experiments and numerical simulations of disordered materials under cyclic loading have unveiled puzzling properties such as the convergence to reversible plastic cycles or the possibility to record and read a past state of deformation. We discuss recent results about such memory effects obtained in atomistic simulations and in lattice-based elastoplastic models of amorphous solids.

Linear algebraic primitives are at the core of many modern algorithms in engineering, science, and machine learning. Hence, accelerating these primitives with novel computing hardware would have tremendous economic impact. I'll discuss how a variety of linear algebra problems can be solved by sampling from the thermodynamic equilibrium distribution of a collection of coupled harmonic oscillators.

A common motif amongst peripheral membrane-binding proteins is a disordered polypeptide domain that can be induced into an amphipathic helix (whereby polar and hydrophobic residues segregate to opposing helix surfaces) with a cognate membrane, thereby controlling protein subcellular localization. However, these peptides are often difficult to characterize in isolate, as they are prone to aggregation and methods to measure binding are low throughput. Herein, we present an engineered platform to enable high-throughput characterization of peptide binding to membranes via fluorescence anisotropy that has been cross-validated with corresponding tryptophan fluorescence measurements. These results represent the first steps in a highly scalable program to not only understand the ability of peptides to detect membrane composition but synthesize novel motifs for subcellular localization of therapeutics.

Patchy particles have become a standard model in soft matter physics to investigate complex molecular behavior, from proteins to large colloidal systems. Synthetic colloidal particles with specific directional interactions can act as a playground to deeper understand their molecular counterparts (such as proteins and smaller molecules), but also open up avenues in the design of novel materials, and even mimic active, living matter. A particularly sensitive way of experimentally controlling the attraction between the patches is by critical Casimir interactions, which allows colloids to assemble into various superstructures, such as chains and networks. To understand and explore the behavior of these Casimir systems, we developed a quantitatively accurate potential model. Using these optimized potentials in large-scale simulations we can predict the phase behavior of mixtures of patchy particles, understand the relaxation behavior of colloidal molecules, and explain the experimentally observed anomalous excess of monomers. Next, we take the system out of equilibrium by including self-propelled active particles, causing rings and chains in the network to undergo breakage and rearrangement. Experiments and simulation show that even a low activity already induces dramatic changes into the dynamical behaviour of these colloidal networks. Such activated viscoelastic architectures can possibly act as model for understanding the behavior of living matter.

My lab pursues research directions that provide insight into the emergence, evolution, and stability of complex social and biological systems. In the past, we addressed questions concerning the structure of food webs, whether there are personality types, why some human genes are ignored by scientists, or how to help clinicians avoid errors of omission. Some of those research questions were prompted by their intrinsic value (the good), others because perhaps no one else would pursue them (the bad), and others because we felt the obligation to do so (the ugly). In this talk, I hope to cover three unpublished stories that illustrate those types of choices and the methods we use in our research. The first story asks whether electronic medical records contain information enabling the discovery of clinical states of critical care patients. The second story ask whether we have developed appropriate and reproducible methods for the analysis of CHIP-seq data that prevent cherry picking. The third story investigates the extraordinary increase in the amount of fraudulent publications and whether they pose a serious risk to the scientific enterprise.

Living systems sense their physical environment and process this information to interact back with the environment. Physics plays a key role in this sensorimotor loop by imposing constraints on all of its basic elements, and long-standing efforts in biological physics have proved fruitful in highlighting these physical constraints across organismal scales. An enduring challenge is to explain the diversity of behaviors we observe in nature, which are generated by deceptively simple "learning rules": for e.g., natural selection and associative learning. I will present recent work on rodent navigation that shows how physics-inspired approaches can help explain emergent phenomena in biological learning.

Essential molecules of life—amino acids, nucleic acids, and sugars—are chiral; they exist in mirror-symmetrical pairs. However, biological systems exclusively use only one form of these pairs: right-handed sugars and nucleic acids, along with left-handed amino acids. This phenomenon characterizes life as homochiral. However, the origins of this asymmetry remain elusive, and it is this long-standing mystery that we address in our work. The chiral-induced spin selectivity (CISS) effect has established a strong coupling between electron spin and molecular chirality, and this coupling paves the way for breaking the chiral molecular symmetry by spin-selective processes. Achiral magnetic surfaces, when spin-polarized, can function as chiral agents due to the CISS effect, serving as templates for the asymmetric crystallization of chiral molecules. We studied the spin-selective crystallization of racemic ribo-aminooxazoline (RAO), a central precursor of RNA, on magnetite surfaces—achieving homochirality in two crystallization steps [1]. Moreover, we have shown the chirality-induced avalanche magnetization of magnetite by RAO molecules, which verifies the reciprocal nature of the effect and allows for a cooperative feedback between chiral molecules and magnetic surfaces [2]. Finally, based on empirical evidence, we propose a pathway through which the achieved homochirality in a single chiral compound, RAO, can efficiently propagate throughout the entire prebiotic network, starting from D-nucleic acids, to L-peptides, and then to homochiral metabolites [3]. Our results demonstrate a prebiotically plausible way of achieving systems-level homochirality from completely racemic starting materials.

References:

[1] S. F. Ozturk, Z. Liu, J. D. Sutherland, D. D. Sasselov, Science Advances 2023, 9, eadg8274.

[2] S. F. Ozturk, D. K. Bhowmick, Y. Kapon, Y. Sang, A. Kumar, Y. Paltiel, R. Naaman, D. D. Sasselov, Nature Communications 2023, 14, 6351.

[3] S. F. Ozturk, D. D. Sasselov, J. D. Sutherland, The Journal of Chemical Physics 2023, 159, 061102.

Developmental disorders are severely understudied, in spite of their alarmingly high incidence, with 1 in 6 U.S. children having one or more disabilities or developmental delays. The main challenge is the design of statistically powered studies that can disentangle numerous genetic and environmental factors. We have been working towards addressing this challenge for the developmental abnormalities associated with the germline mutations within the ERK cascade. Focusing on mutations that affect MEK, a kinase that activates ERK, we demonstrated how studies of human mutations in Drosophila can answer the long-standing questions in the field. Specifically, we established how pathogenic mutations affect an isolated MEK protein, demonstrated how they disrupt the normal process of MEK activation in the cell, and quantified their effects on ERK signaling in embryos. More recently, our work shed light on the origins of phenotypic variability in the ERK-associated developmental disorders, demonstrating that they can be of purely stochastic origin. Given the generality of our approach, it should be applicable to other developmental abnormalities associated with genetically deregulated cell signaling.

Many biological systems operate near the physical limits to their performance, suggesting that aspects of their behavior and underlying mechanisms could be derived from optimization principles. However, such principles have often been applied only in simplified models. Here, we explore a detailed mechanistic model of the gap gene network in the Drosophila embryo, optimizing its 50+ parameters to maximize the information that gene expression levels provide about nuclear positions. This optimization is conducted under realistic constraints, such as limits on the number of available molecules. Remarkably, the optimal networks we derive closely match the architecture and spatial gene expression profiles observed in the real organism. Our framework quantifies the trade-offs involved in maximizing functional performance and allows for the exploration of alternative network configurations, addressing the question of which features are necessary and which are contingent. Our results suggest that multiple solutions to the optimization problem might exist across closely related organisms, offering new insights into the evolution of gene regulatory networks.

In this talk I will discuss algorithms that learn to predict a target’s location using its odor, and navigate to reach it despite sparsity induced by turbulence. First, I will use supervised learning to show that sensing for as little as a few seconds is sufficient to predict where an olfactory target is. Robust predictions are achieved by combining complementary information from intensity and sparsity of the odor. Inspired by octopus ambush predation, I will show that noise unintuitively improves predictions. Second, I will address navigation, motivated by behavioral experiments in sea robins — fish with legs that are used to facilitate walking and digging along the sea floor. I will discuss different options for turbulent navigation, ranging from planning to learning. I will show that navigation can be either learned as a stimulus-response function through reinforcement learning, or it can be planned using prior information on the odor plume. Learning can be achieved with a short temporal memory and an optimal memory emerges due to the need to negotiate between following the odor within the odor plume and recovering the plume after a inadvertently exiting. Memory connects to the sparsity of the odor dictated by turbulence, suggesting physics can be used to set memory for olfactory searches.

The origin of life on Earth is a profound mystery for which scientific answers are lacking. I will report on laboratory experiments that attempt to recreate molecular evolution under the conditions of the early Earth. They mimic the geophysical nonequilibria on a volcanic island and combine them with an early biochemistry that could form and replicate RNA sequences. The environments we study include wet-dry cycles, isothermal airflow, and systems driven by temperature differences, both with an air-water interface [1] and without [2]. In addition to using proteins to accelerate evolutionary dynamics, we are attempting to initiate early evolution from activated 2',3'-cyclic nucleotides in the alkaline environment of volcanic rock. This includes the formation of RNA, its copying by templated ligation [3], evidence for early molecular cooperation with amino acids, and the formation of protocells. Our goal is to find geochemical environments that bridge the early evolution of RNA to the formation of the first cells.

[1] Ianeselli et. al., Nature Physics doi.org/10.1038/s41567-022-01516-z (2022)

[2] Matreux et. al., Nature doi.org/10.1038/s41586-024-07193-7 (2024)

[3] Serrão et. al., JACS doi.org/10.1021/jacs.3c10813 (2024)

From swarms of swimming bacteria to the moving contents of cells, biology is replete with active systems whose microscopic constituents interact by performing mechanical work on a surrounding fluidic medium. This can lead to large-scale, sometimes functional, self-organized structures and complex dynamics. I'll overview the modeling of such systems, focusing first on continuum kinetic theories that couple the micro and macroscopic scales to describe how suspensions of active particles, such as swimming microorganisms, evolve in time. While high-dimensional (5+1) these models have been used to understand observations of novel instabilities, turbulent-like dynamics, and strange rheology, and have been incorporated into more complex models of biological systems. I'll then pivot to describe the emergence of large scale, spontaneously appearing transport flows in developing egg cells. Building on a conception of molecular motors carrying payloads on a flexible polymer assembly, I'll develop an active porous medium model whose instabilities naturally drive the system towards large-scale "twister" flows consistent with experiments.

Superconducting resonators are technological building blocks for experiments in quantum computing, cosmology, and particle physics. Yet, despite their prevalence, in some limits they can exhibit rich and poorly understood behavior. Resonators formed from an array of Josephson junctions are a prime example. I will present two studies exploring their physics. The first study shows that apparent superconductivity persists for vastly weaker arrays than expected within a zero-temperature theory. This behavior is consistent with thermal effects, which effectively melt the insulator and restore superconducting behavior [1]. The second study explores a source of dissipation arising from photon-photon interactions — photonic “friction”. I will discuss our current efforts to characterize both decay rates and kinetics associated with this effect.

[1] S. Mukhopadhyay et al., Nat. Phys. 19 (2023) 1630.

Cells express genes when they respond to environmental changes, differentiate to different cell fates, or develop into a healthy organism. Gene expression is often regulated by externally supplied cues, such as changing transcription factor concentrations. The expression in response to a changing concentration can be viewed as a type of decision that can be analyzed in terms of an information-theoretic framework. In this talk, I will show, on the examples of early development in the fruit fly and in cultured mouse stem cells, how an information-theoretic inference approach can help us understand features of a complex signalling apparatus that may be difficult to model, due to the complexity of the contributing regulatory factors. One inference approach, the information bottleneck, simplifies for molecular sensing, where signals are smooth; I will show how this can be used to understand binding site architectures. Finally, I will discuss our work on modelling and inference on noisy molecular signals in the context of wnt signalling.

Accurate and efficient nonlinear dynamical systems models are essential to understand, predict, estimate, and control complex natural and engineered systems. In this talk, I will explore how machine learning may be used to develop these models purely from measurement data. We explore the sparse identification of nonlinear dynamics (SINDy) algorithm, which identifies a minimal dynamical system model that balances model complexity with accuracy, avoiding overfitting. This approach tends to promote models that are interpretable and generalizable, capturing the essential “physics” of the system. We also discuss the importance of learning effective coordinate systems in which the dynamics may be expected to be sparse. This sparse modeling approach will be demonstrated on a range of challenging modeling problems, for example in fluid dynamics. Because fluid dynamics is central to transportation, health, and defense systems, we will emphasize the importance of machine learning solutions that are interpretable, explainable, generalizable, and that respect known physics.