Bulletin of the American Physical Society
APS March Meeting 2020
Volume 65, Number 1
Monday–Friday, March 2–6, 2020; Denver, Colorado
Session R22: Inference, Information, and Learning in Biophysics: IIFocus
|
Hide Abstracts |
Sponsoring Units: DBIO GSNP DCOMP Chair: Paul Wiggins Room: 303 |
Thursday, March 5, 2020 8:00AM - 8:36AM |
R22.00001: The strange case of Dr Jekyll and Mr Hyde: The two faces of singular models. Invited Speaker: Paul Wiggins Why does systems biology work in spite of a blizzard of poorly-defined parameters and yet the detection of the Higgs boson requires five-sigma? Both these statistical analyses involve singular models, defined by structural unidentifiability (i.e. the absence of a one-to-one map between parameters and distribution functions). This singular structure leads to profound changes in the phenomenology of inference. In this talk, we will explore the phenomenology of learning from two physical perspectives: First, we explore the correspondence between statistical physics and statistics and demonstrate that there is equivalence between predictive performance and heat capacity, which gives new physical insight into why learning has universal scaling as well as explaining how and why these universal rules fail in the context of singular models. Finally, we explore insights from the Riemannian geometry of the model parameter space to determine what face a singular model will show: anomalously high or low learning performance. |
Thursday, March 5, 2020 8:36AM - 8:48AM |
R22.00002: Learning dynamical information from static protein and sequencing data Philip Pearce, Francis G Woodhouse, Aden W Forrow, Ashley Kelly, Halim Kusumaatmaja, Jorn Dunkel Many complex processes, from protein folding to neuronal network dynamics, can be described as stochastic exploration of a high-dimensional energy landscape. While efficient algorithms for cluster detection in high-dimensional spaces have been developed over the last two decades, considerably less is known about the reliable inference of state transition dynamics in such settings. Here, we introduce a flexible and robust numerical framework to infer Markovian transition networks directly from time-independent data sampled from stationary equilibrium distributions. We demonstrate the practical potential of the inference scheme by reconstructing the network dynamics for several protein folding transitions, gene-regulatory network motifs and HIV evolution pathways. The predicted network topologies and relative transition time scales agree well with direct estimates from time-dependent molecular dynamics data, stochastic simulations and phylogenetic trees, respectively. Owing to its generic structure, the framework introduced here will be applicable to high-throughput RNA and protein sequencing datasets and future cryo-electron-microscopy data. |
Thursday, March 5, 2020 8:48AM - 9:00AM |
R22.00003: Associative Memory of Structured Knowledge Julia Steinberg, Haim Sompolinsky A long standing challenge in biological and artificial intelligence is to understand how new knowledge can be constructed from known building blocks in a way that is amenable for computation by neuronal circuits. While previous work has focused primarily on working memory tasks of structured data, here we focus on the task of storage and recall of structured knowledge in long term memory. Specifically, we ask how a Hopfield type network can store and retrieve episodic memories where each episode is a set of associations between events. We model each knowledge structure as a set of binary relations between events and cues (cues may represent e.g., temporal order, spatial location, role in semantic structure). We use a binarized version of holographic reduced representation (HRR) to map such structures to fixed length vectors. We then train a recurrent network to store these vectors as fixed points. By a combination of signal-to-noise analysis and numerical simulations we demonstrate that our model allows for an efficient storage and recall of these knowledge structures in a way that allows the full retrieval of the memorized structure and their building blocks from partial retrieving cues. Our work contributes to the understanding of neural computations of structured knowledge. |
Thursday, March 5, 2020 9:00AM - 9:12AM |
R22.00004: Can model reduction replace expert intuition for modeling complex biological systems? Cody Petrie, Dane Bjork, Mark Transtrum One of the challenges to modeling biological systems is the overwhelming complexity. Mathematical models that account for all the known interactions would have an unwieldy number of components and parameters. Traditionally, real-world models have been based on expert intuition that judiciously relate only those components believed to be relevant to a behavior of interest. These models typically reflect intuition and physical insights that are difficult to rigorously justify or convey. Here, we consider whether recent advances in model reduction may be able to automatically construct comparable simplified models in a semi-automatic, data-driven way. We report on a comparative study of model reduction of the Wnt signaling pathway, comparing automatic methods with expert intuition. Automatic model reduction is done using the Manifold Boundary Approximation Method (MBAM), based on information geometry and "sloppy model" analysis. We find that MBAM leads to simplified models that closely resemble those proposed based on expert intuition. Our results suggest that data-driven methods of model reduction may be a viable alternative to expert-derived models, and can be used to extract comparable physical insights into the behavior of complex biological systems. |
Thursday, March 5, 2020 9:12AM - 9:24AM |
R22.00005: Limits on the suppression of molecular fluctuations and oscillation dephasing in stochastic reaction networks Jiawei Yan, Johan Paulsson Many efforts in synthetic biology have been dedicated to designing ultra-reliable gene networks, but so far there has been little theory capable of providing guidance on the experimental design, because small differences in rate functions or topology can sometimes change the dynamics drastically. Here we aim to identify general principles in stochastic reaction networks that apply regardless of parameters and the form of rate functions. First we studied the noise suppression in multi-component networks and asked if it is possible to design systems where the components control and mutually suppress the noise in each other. Specifically, we find that in any N-component system, regardless of how each component affects other's production rates, it is impossible to suppress fluctuations below the uncontrolled equivalents for all components. We next examined whether there exist similar design principles in oscillatory behaviours in stochastic reaction networks. We studied a broad class of feedback, allowing arbitrary time delay and control functions, and found that even when all the rest of the feedback loop is optimal for generating sustained oscillations, the information loss from one single reaction step can lead to severe constraints in the autocorrelation and power spectrum. |
Thursday, March 5, 2020 9:24AM - 9:36AM |
R22.00006: Kalman-like Self-Tuned Sensitivity in Biophysical Sensing Kabir Husain, Weerapat Pittayakanchit, Gopal Pattanayak, Michael Rust, Arvind Murugan Living organisms need to be sensitive to a changing environment while also ignoring uninformative environmental fluctuations. Here, we argue that living cells can navigate these conflicting demands by dynamically tuning their environmental sensitivity. We analyze the circadian clock in Synechococcus elongatus, showing that clock-metabolism coupling can detect mismatch between clock predictions and the day-night light cycle, temporarily raise the clock’s sensitivity to light changes, and thus re-entraining faster. We find analogous behavior in recent experiments on switching between slow and fast osmotic-stress-response pathways in yeast. In both cases, cells can raise their sensitivity to new external information in epochs of frequent challenging stress, much like a Kalman filter with adaptive gain in signal processing. Our work suggests a new class of experiments that probe the history dependence of environmental sensitivity in biophysical sensing mechanisms. |
Thursday, March 5, 2020 9:36AM - 9:48AM |
R22.00007: Searching for the Relevant Properties of Binary Datasets: Is your Model Truly Pairwise? Clelia De Mulatier, Paolo Pietro Mazza, Matteo Marsili Uncovering the patterns hidden within noisy data is essential to science. Information theory provides a quantitative method to select the best of potential explanations for data, by optimizing the balance between goodness-of-fit and simplicity. Yet in practice finding “the” best model for a given dataset is impossible. A common practical issue is the huge number of potential models. But with a finite amount of data, the real limitation comes from the large degeneracy of models that perform nearly optimally. We illustrate this problem on examples of binary data using a heuristic procedure to perform an efficient search among all spin models with high order interactions. As good models tend to share a common sub-structure that is likely to capture relevant properties of the data, we focus our search on this structure rather than on finding the strictly best model. We show that minimally complex spin models are useful for this task. We obtain an analytic expression for their posterior probability, which makes them easy to fit and exactly comparable. We then show that working with equivalence classes of these models allows a) to find the spin basis in which the dependencies between basis variables are minimal and b) to quantify these dependencies and the relevance of each dimension. |
Thursday, March 5, 2020 9:48AM - 10:00AM |
R22.00008: Information efficiency of bacterial chemotaxis Henry Mattingly, Keita Kamino, Xiaowei Zhang, Benjamin B Machta, Thierry Emonet Information transfer is central to the function of many biological systems. For example, the bacteria Escherichia coli climbs gradients of chemical attractants by modulating its rate of tumbling—randomly reorienting its swimming direction—when it senses time-changes in attractant concentration. But even in the absence of signal processing, the cell’s tumble behavior is correlated with the signal it sees. The transfer entropy rate Iφ→M from signal to tumble behavior removes these correlations and isolates the causal influence of the signal on the cell’s tumble decisions. We show that climbing a gradient with drift speed vD requires an information rate of at least Iφ→M = 12 Dr (vD/v0)2 (1-TB), where Dr is the rate of rotational diffusion, v0 is the run speed, and TB is the fraction of time the cell is tumbling. To quantify E. coli cells' information rates, we measure, in single cells, signal transduction responses and fluctuations. Along with measurements of drift speeds, we determine how efficiently E. coli use information about the gradient during chemotaxis. |
Thursday, March 5, 2020 10:00AM - 10:12AM |
R22.00009: Decision-making at a T-junction by gradient-sensing agents Tanvi Gandhi, Jinzi Mac Huang, Antoine Aubret, Desmond Yaocheng Li, Sophie Ramananarivo, Massimo Vergassola, Jeremie Palacci At the beginning of life, searching for food and evading hazards are two essential activities for microorganisms to survive, and the way they navigate is through chemotaxis. The optimal chemotaxis in complicated terrains determines the fate of living creatures, and natural selection ensures the existence of such an optimization. In our study, we investigate the navigation of inert particles in a network that has multiple junctions. In micro-networks manufactured through photolithography, a background gradient of salt is established as the signal of chemoattractant by placing a source and a sink of salt. Colloidal particles then follow this signal through diffusiophoresis and move towards the source. Through stochastic modeling, we show that particles prefer to exit each junction at the end with higher concentration gradient. This preference is further enhanced when the particle size is larger, which leads to a way to magnify small signals in a network so that the colloidal particles larger than a critical size can always move towards the source of salt through the shortest path. Ultimately, we compare the navigation schemes of inert particles and living organisms, aiming to understand biological chemotaxis and shed light on future manufacturing of navigable microswimmers. |
Thursday, March 5, 2020 10:12AM - 10:24AM |
R22.00010: Trading bits in the readout of positional information Marianne Bauer, William S Bialek, Thomas Gregor, Mariela D Petkova, Eric Wieschaus Expression levels of gap genes, crucial for fly development, have been shown to be precise, matching the spatial precision in the expression of downstream genes. If the transcriptional machinery can read these gap expression levels only imprecisely, subsequent precision could not be reproduced. Yet, assuming that levels can be read with infinite precision is unrealistic. Reading expression levels with limited precision can be phrased mathematically by limiting the number of bits which are available for each measurement. The question of how to use these bits to capture the maximum positional information is an instance of the information bottleneck problem. We show that to capture ~90% of the available information, we need more bits than intuitively biologically reasonable. We can allow for an increased number of bits per measurement by having molecules bind to multiple sites. This is a generalization of the information bottleneck problem. We show that one can capture almost all the available information with multiple “low precision” encodings, which may correspond physically to multiple binding sites or even enhancers. |
Thursday, March 5, 2020 10:24AM - 10:36AM |
R22.00011: Single-Molecule Conductance and Conformational Analysis with Engineered Nano-Junctions for Nucleic Acid Sequencing Lee Korshoj, Sepideh Afsari, Anushree Chatterjee, Prashant Nagpal DNA sequencing on the single-molecule level can be used to study cellular heterogeneity and stochasticity with reduced time, cost, and complexity compared to traditional sequencing methods. However, sample noise and signature overlap due to varying nucleotide conformations prevent accurate sequencing results. We address these issues by engineering nano-junctions for conductance measurements on conformationally constrained single nucleotides within electrostatically bound DNA molecules on a self-assembled cysteamine monolayer. From STM break junctions with biochemical moieties in individual nucleobases, the unique conductance signature of each nucleobase is analyzed with machine learning algorithms [1]. Additionally, conformational variation, or smear, is quantified from the distance over which molecular junctions are maintained during each conductance measurement [2]. We demonstrate >93% accuracy for DNA nucleotide recognition with 20 repeat measurements. These results are a significant improvement over contemporary methods and show the potential for using simple surface modifications and existing biochemical moieties in nucleobases for single-molecule, nanoelectronic nucleotide identification. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700