Qiwei Yu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Qiwei Yu has a total of 70 co-authors. Here are the top co-authors by number of shared publications:
- Yuhai Tu
Center for Computational Biology and Center for Computational Neuroscience, Flatiron Institute, New York, NY, USA
6 shared publications - Andrej Košmrlj
Princeton University
5 shared publications - Ned S. Wingreen
Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, NJ 08544, USA
5 shared publications - Aya A. Abu-Alfa
Department of Molecular Biology, Princeton University, Princeton, NJ 08544, USA
3 shared publications - Clifford P. Brangwynne
Princeton University, USA
3 shared publications - Jordy F. Botello
Department of Molecular Biology, Princeton University, Princeton, NJ 08544, USA
3 shared publications
Latest Publications
- Open Access
- 10 November 2026
- Published by Springer Nature in Nature Microbiology
- p. 1-11
S-layer expansion is mediated by hoop-like cracks in the paracrystalline lattice during Bacillus anthracis growth
Surface layers (S-layers) are paracrystalline protein monolayers that envelop archaea and some bacteria, including Bacillus anthracis. S-layer subunits in B. anthracis are tethered to the underlying cell wall through glycopolymers that are covalently attached to the peptidoglycan (PG) meshwork. Cell growth in this rod-shaped organism occurs by dispersed PG synthesis throughout the cell cylinder. How the crystalline lattice expands in lockstep with the cell wall has remained unclear. Here, using fluorescently labelled S-layer subunits and high-resolution imaging, we show that S-layer expansion during B. anthracis growth is mediated by pervasive hoop-like cracks in the lattice. Pulse-chase experiments reveal two modes of S-layer assembly at these cracks: dispersed and focused. Our data suggest that local pulling forces on the S-layer array generated by dispersed cell wall synthesis and expansion of the underlying PG cause crack formation, with ensuing dispersed and focused synthesis enabling S-layer expansion with bacterial growth.
- 10 November 2026
- Published by Springer Nature in Nature Microbiology
- p. 1-11
Open AccessS-layer expansion is mediated by hoop-like cracks in the paracrystalline lattice during Bacillus anthracis growth
Surface layers (S-layers) are paracrystalline protein monolayers that envelop archaea and some bacteria, including Bacillus anthracis. S-layer subunits in B. anthracis are tethered to the underlying cell wall through glycopolymers that are covalently attached to the peptidoglycan (PG) meshwork. Cell growth in this rod-shaped organism occurs by dispersed PG synthesis throughout the cell cylinder. How the crystalline lattice expands in lockstep with the cell wall has remained unclear. Here, using fluorescently labelled S-layer subunits and high-resolution imaging, we show that S-layer expansion during B. anthracis growth is mediated by pervasive hoop-like cracks in the lattice. Pulse-chase experiments reveal two modes of S-layer assembly at these cracks: dispersed and focused. Our data suggest that local pulling forces on the S-layer array generated by dispersed cell wall synthesis and expansion of the underlying PG cause crack formation, with ensuing dispersed and focused synthesis enabling S-layer expansion with bacterial growth.
- Open Access
A minimal kinase-phosphatase system and its lipid substrates self-organize into dynamic patterns
Competing lipid kinases and phosphatases are critical for organizing cellular membranes, but whether a minimal system can autonomously organize proteins and lipids into dynamic spatiotemporal patterns is unknown. Here, we report the in vitro reconstitution of the Legionella phosphatidylinositol (PI) 3-kinase MavQ and PI 3-phosphatase SidP. Together with their lipid substrates, PI and PI 3-phosphate (PI3P), these enzymes form a minimal self-organizing system that generates ATP-dependent spatiotemporal patterns, including traveling waves, on model membranes. These behaviors arise from MavQ's cooperative membrane binding, SidP's phosphatase activity, and the continual interconversion and redistribution of PI and PI3P within a conserved membrane pool. A reaction-diffusion model reproduces the observed dynamics and predicts that lipid conservation prevents patterns from propagating across membrane discontinuities, which we verify experimentally. Together, these findings establish enzymatic modification of membrane lipids as a distinct molecular strategy for biological pattern formation.
Open AccessA minimal kinase-phosphatase system and its lipid substrates self-organize into dynamic patterns
Competing lipid kinases and phosphatases are critical for organizing cellular membranes, but whether a minimal system can autonomously organize proteins and lipids into dynamic spatiotemporal patterns is unknown. Here, we report the in vitro reconstitution of the Legionella phosphatidylinositol (PI) 3-kinase MavQ and PI 3-phosphatase SidP. Together with their lipid substrates, PI and PI 3-phosphate (PI3P), these enzymes form a minimal self-organizing system that generates ATP-dependent spatiotemporal patterns, including traveling waves, on model membranes. These behaviors arise from MavQ's cooperative membrane binding, SidP's phosphatase activity, and the continual interconversion and redistribution of PI and PI3P within a conserved membrane pool. A reaction-diffusion model reproduces the observed dynamics and predicts that lipid conservation prevents patterns from propagating across membrane discontinuities, which we verify experimentally. Together, these findings establish enzymatic modification of membrane lipids as a distinct molecular strategy for biological pattern formation.
- Open Access
- 15 May 2026
- Published by American Physical Society (APS) in PRX Life
- Vol. 4 (2)
Membrane-Anchored Mobile Tethers Modulate Condensate Wetting, Localization, and Migration
Biomolecular condensates frequently rely on membrane interactions for recruitment, localization, and biochemical substrates. Many of these interactions are mediated by membrane-anchored molecules such as proteins or specific lipids, which we refer to as “mobile tethers” since they can typically diffuse within the membrane while still interacting with the condensate. The presence of mobile tethers creates a surface with dynamic and spatially inhomogeneous wetting properties that are typically overlooked by traditional wetting theories. Here, we propose a general theoretical framework to study how mobile tethers impact both equilibrium and dynamic properties of condensate wetting. We show that a favorable tether-condensate interaction leads to tether enrichment at the condensate-membrane interface, which modifies the equilibrium condensate–membrane surface tension and contact angle. Increasing tether abundance on the membrane can drive transitions between wetting regimes, with only a modest tether density and binding energy required for biologically relevant scenarios. Furthermore, tethers modulate how condensates react to complex membrane geometries. By helping condensates coat membranes, mobile tethers can facilitate condensate localization to junctions of membrane structures, such as the reticulated membranes inside the algal pyrenoid. Both tether abundance and mobility affect how droplets interact with complex membrane geometries, such as droplet migration along membrane tubules of varying radii. These results provide a framework to study the implications of tether-mediated condensate-membrane interactions for cellular organization and function.
- 15 May 2026
- Published by American Physical Society (APS) in PRX Life
- Vol. 4 (2)
Open AccessMembrane-Anchored Mobile Tethers Modulate Condensate Wetting, Localization, and Migration
Biomolecular condensates frequently rely on membrane interactions for recruitment, localization, and biochemical substrates. Many of these interactions are mediated by membrane-anchored molecules such as proteins or specific lipids, which we refer to as “mobile tethers” since they can typically diffuse within the membrane while still interacting with the condensate. The presence of mobile tethers creates a surface with dynamic and spatially inhomogeneous wetting properties that are typically overlooked by traditional wetting theories. Here, we propose a general theoretical framework to study how mobile tethers impact both equilibrium and dynamic properties of condensate wetting. We show that a favorable tether-condensate interaction leads to tether enrichment at the condensate-membrane interface, which modifies the equilibrium condensate–membrane surface tension and contact angle. Increasing tether abundance on the membrane can drive transitions between wetting regimes, with only a modest tether density and binding energy required for biologically relevant scenarios. Furthermore, tethers modulate how condensates react to complex membrane geometries. By helping condensates coat membranes, mobile tethers can facilitate condensate localization to junctions of membrane structures, such as the reticulated membranes inside the algal pyrenoid. Both tether abundance and mobility affect how droplets interact with complex membrane geometries, such as droplet migration along membrane tubules of varying radii. These results provide a framework to study the implications of tether-mediated condensate-membrane interactions for cellular organization and function.
- Open Access
Ribosome biogenesis bottlenecks reveal vulnerabilities in cancer
Summary: Cell growth requires elevated protein synthesis, which depends on the production of ribosomes. Ribosome biogenesis is a complex, multi-step pathway in which newly transcribed precursor ribosomal RNA (rRNA) undergoes coordinated processing and assembly in the nucleolus to produce the small and large ribosomal subunits (SSU and LSU). 1–3 Oncogene activation stimulates rRNA transcription and processing, giving rise to enlarged nucleoli that produce thousands of ribosomes every minute. 4,5 However, efficient ribosome production requires tight coordination across numerous maturation steps, and it remains unclear if elevated rDNA transcription is proportionally converted into mature ribosomes, or whether imperfect coordination constrains the output yield. Here, we quantify pre-rRNA transcription (input) and compare it with newly-assembled cytoplasmic ribosomes (output), revealing that oncogene activation reduces the efficiency of ribosome production. Using a quantitative pulse-chase sequencing approach with mathematical modeling to resolve rRNA maturation kinetics, we found that oncogene activation creates late-stage processing bottlenecks, characterized by delayed precursor maturation and increased degradation. Perturbation of late-stage ribosome biogenesis factors preferentially impaired oncogene-driven cell growth, and limited tumor growth in mouse models, suggesting that these bottlenecks represent selective vulnerabilities in cancer, created by imbalanced biosynthetic flux. Together, these findings reveal that oncogene-driven ribosome production is imperfectly coordinated across maturation steps, and suggest that capacity limits in multi-step assembly pathways may be therapeutically exploitable in cancer and other diseases.
Open AccessRibosome biogenesis bottlenecks reveal vulnerabilities in cancer
Summary: Cell growth requires elevated protein synthesis, which depends on the production of ribosomes. Ribosome biogenesis is a complex, multi-step pathway in which newly transcribed precursor ribosomal RNA (rRNA) undergoes coordinated processing and assembly in the nucleolus to produce the small and large ribosomal subunits (SSU and LSU). 1–3 Oncogene activation stimulates rRNA transcription and processing, giving rise to enlarged nucleoli that produce thousands of ribosomes every minute. 4,5 However, efficient ribosome production requires tight coordination across numerous maturation steps, and it remains unclear if elevated rDNA transcription is proportionally converted into mature ribosomes, or whether imperfect coordination constrains the output yield. Here, we quantify pre-rRNA transcription (input) and compare it with newly-assembled cytoplasmic ribosomes (output), revealing that oncogene activation reduces the efficiency of ribosome production. Using a quantitative pulse-chase sequencing approach with mathematical modeling to resolve rRNA maturation kinetics, we found that oncogene activation creates late-stage processing bottlenecks, characterized by delayed precursor maturation and increased degradation. Perturbation of late-stage ribosome biogenesis factors preferentially impaired oncogene-driven cell growth, and limited tumor growth in mouse models, suggesting that these bottlenecks represent selective vulnerabilities in cancer, created by imbalanced biosynthetic flux. Together, these findings reveal that oncogene-driven ribosome production is imperfectly coordinated across maturation steps, and suggest that capacity limits in multi-step assembly pathways may be therapeutically exploitable in cancer and other diseases.
- Open Access
Ribosome Molecular Aging Shapes Translation Dynamics
SUMMARY: Cellular homeostasis relies on continual renewal of cellular components, yet some complexes like ribosomes persist for long periods, raising the question of whether extended molecular age impacts functional fidelity. Here, we introduce a spatiotemporal mapping strategy to resolve biomolecular life stages, and show that intracellular ribosome aging alters translational dynamics at specific transcripts. Molecularly aged ribosomes exhibit impaired elongation at basic amino acid-rich sequences, leading to increased pausing, premature termination, and ribosome collisions. By profiling ribosomal RNA modifications, we find that molecular aging increases the collision propensity of specific ribosome subpopulations. Consistent with our findings, enrichment of aged ribosomes in cells amplifies molecular age-dependent translation defects. In vivo labeling of ribosomes in aged C. elegans demonstrates that molecularly aged ribosomes shape translational dynamics during organismal aging. These findings identify ribosome molecular age as a determinant of translational dynamics, and link molecular aging of a core gene-expression complex to organismal aging. HIGHLIGHTS: A pulse-chase labeling strategy enables mapping subcellular demographics of macromolecular complexes in space and time. Molecular aging of ribosomes drives differential mRNA translation and shapes elongation dynamics. The collision propensity of specific ribosome subpopulations increases with molecular age. Older ribosomes shape translation dynamics during organismal aging.
Open AccessRibosome Molecular Aging Shapes Translation Dynamics
SUMMARY: Cellular homeostasis relies on continual renewal of cellular components, yet some complexes like ribosomes persist for long periods, raising the question of whether extended molecular age impacts functional fidelity. Here, we introduce a spatiotemporal mapping strategy to resolve biomolecular life stages, and show that intracellular ribosome aging alters translational dynamics at specific transcripts. Molecularly aged ribosomes exhibit impaired elongation at basic amino acid-rich sequences, leading to increased pausing, premature termination, and ribosome collisions. By profiling ribosomal RNA modifications, we find that molecular aging increases the collision propensity of specific ribosome subpopulations. Consistent with our findings, enrichment of aged ribosomes in cells amplifies molecular age-dependent translation defects. In vivo labeling of ribosomes in aged C. elegans demonstrates that molecularly aged ribosomes shape translational dynamics during organismal aging. These findings identify ribosome molecular age as a determinant of translational dynamics, and link molecular aging of a core gene-expression complex to organismal aging. HIGHLIGHTS: A pulse-chase labeling strategy enables mapping subcellular demographics of macromolecular complexes in space and time. Molecular aging of ribosomes drives differential mRNA translation and shapes elongation dynamics. The collision propensity of specific ribosome subpopulations increases with molecular age. Older ribosomes shape translation dynamics during organismal aging.
- Open Access
- 24 July 2025
- Published by American Physical Society (APS) in Physical Review Letters
- Vol. 135 (4) , 048402
Evolutionary Benefits of Fitness-Dependent Mutation Rates
Motivated by recent observations that mutation rates can be correlated with individual fitness, we analyze an evolutionary hill-climbing model with fitness-dependent mutation rates. Our results show that a mutation rate that decreases with increasing relative fitness can greatly accelerate the accumulation of beneficial mutations. Moreover, we show that a lower mutation rate for fitter individuals can prevent “mutational meltdown” of small populations by decreasing the probability of fixation of deleterious mutations. These findings suggest potential strategies for accelerating the adaptation of populations to environmental changes.
- 24 July 2025
- Published by American Physical Society (APS) in Physical Review Letters
- Vol. 135 (4) , 048402
Open AccessEvolutionary Benefits of Fitness-Dependent Mutation Rates
Motivated by recent observations that mutation rates can be correlated with individual fitness, we analyze an evolutionary hill-climbing model with fitness-dependent mutation rates. Our results show that a mutation rate that decreases with increasing relative fitness can greatly accelerate the accumulation of beneficial mutations. Moreover, we show that a lower mutation rate for fitter individuals can prevent “mutational meltdown” of small populations by decreasing the probability of fixation of deleterious mutations. These findings suggest potential strategies for accelerating the adaptation of populations to environmental changes.
- Top Cited PapersOpen Access
- 2 July 2025
- Published by Springer Nature in Nature
- Vol. 644 (8076) , 557-566
Mapping and engineering RNA-driven architecture of the multiphase nucleolus
Biomolecular condensates are key features of intracellular compartmentalization1,2. As the most prominent nuclear condensate in eukaryotes, the nucleolus is a multiphase liquid-like structure in which ribosomal RNAs (rRNAs) are transcribed and processed, undergoing multiple maturation steps to form the small (SSU) and large (LSU) ribosomal subunits3,4,5. However, how rRNA processing is coupled to the layered organization of the nucleolus is poorly understood owing to a lack of tools to precisely monitor and perturb nucleolar rRNA processing dynamics. Here we developed two complementary approaches to spatiotemporally map rRNA processing and engineer de novo nucleoli. Using sequencing in parallel with imaging, we found that rRNA processing steps are spatially segregated, with sequential maturation of rRNA required for its outward movement through nucleolar phases. By generating synthetic nucleoli in cells using an engineered rDNA plasmid system, we show that defects in SSU processing can alter the ordering of nucleolar phases, resulting in inside-out nucleoli and preventing rRNA outflux, while LSU precursors are necessary to build the outermost layer of the nucleolus. These findings demonstrate how rRNA is both a scaffold and substrate for the nucleolus, with rRNA acting as a programmable blueprint for the multiphase architecture that facilitates assembly of an essential molecular machine.
- 2 July 2025
- Published by Springer Nature in Nature
- Vol. 644 (8076) , 557-566
Top Cited PapersOpen AccessMapping and engineering RNA-driven architecture of the multiphase nucleolus
Biomolecular condensates are key features of intracellular compartmentalization1,2. As the most prominent nuclear condensate in eukaryotes, the nucleolus is a multiphase liquid-like structure in which ribosomal RNAs (rRNAs) are transcribed and processed, undergoing multiple maturation steps to form the small (SSU) and large (LSU) ribosomal subunits3,4,5. However, how rRNA processing is coupled to the layered organization of the nucleolus is poorly understood owing to a lack of tools to precisely monitor and perturb nucleolar rRNA processing dynamics. Here we developed two complementary approaches to spatiotemporally map rRNA processing and engineer de novo nucleoli. Using sequencing in parallel with imaging, we found that rRNA processing steps are spatially segregated, with sequential maturation of rRNA required for its outward movement through nucleolar phases. By generating synthetic nucleoli in cells using an engineered rDNA plasmid system, we show that defects in SSU processing can alter the ordering of nucleolar phases, resulting in inside-out nucleoli and preventing rRNA outflux, while LSU precursors are necessary to build the outermost layer of the nucleolus. These findings demonstrate how rRNA is both a scaffold and substrate for the nucleolus, with rRNA acting as a programmable blueprint for the multiphase architecture that facilitates assembly of an essential molecular machine.
- Open Access
- 1 July 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 1-16
Neurons exploit stochastic growth to rapidly and economically build dense dendritic arbors
Dendrites grow by stochastic branching, elongation, and retraction. A key question is whether such a mechanism is sufficient to form highly branched dendritic morphologies. Alternatively, does dendrite geometry depend on signals from other cells or from the topological hierarchy of the growing network? To answer these questions, we developed an isotropic and homogenous mean-field model in which branch dynamics depends only on average lengths and densities: that is, without external influence. Branching was modeled as density-dependent nucleation so that no tree structures or network topology was present. Despite its simplicity, the model predicted several key morphological properties of class IV Drosophila sensory dendrites, including the exponential distribution of branch lengths, the parabolic scaling between dendrite number and length densities, the tight spacing of the dendritic meshwork (which required minimal total branch length), and the radial orientation of branches. Stochastic growth also accelerated the overall expansion rate of the arbor. We show that stochastic dynamics is an economical and rapid space-filling mechanism for building dendritic arbors without external guidance or hierarchical branching mechanisms. Our work therefore provides a general theoretical framework for understanding how macroscopic branching patterns emerge from microscopic dynamics.
- 1 July 2025
- Published by Springer Nature in Nature Communications
- Vol. 16 (1) , 1-16
Open AccessNeurons exploit stochastic growth to rapidly and economically build dense dendritic arbors
Dendrites grow by stochastic branching, elongation, and retraction. A key question is whether such a mechanism is sufficient to form highly branched dendritic morphologies. Alternatively, does dendrite geometry depend on signals from other cells or from the topological hierarchy of the growing network? To answer these questions, we developed an isotropic and homogenous mean-field model in which branch dynamics depends only on average lengths and densities: that is, without external influence. Branching was modeled as density-dependent nucleation so that no tree structures or network topology was present. Despite its simplicity, the model predicted several key morphological properties of class IV Drosophila sensory dendrites, including the exponential distribution of branch lengths, the parabolic scaling between dendrite number and length densities, the tight spacing of the dendritic meshwork (which required minimal total branch length), and the radial orientation of branches. Stochastic growth also accelerated the overall expansion rate of the arbor. We show that stochastic dynamics is an economical and rapid space-filling mechanism for building dendritic arbors without external guidance or hierarchical branching mechanisms. Our work therefore provides a general theoretical framework for understanding how macroscopic branching patterns emerge from microscopic dynamics.
- Open Access
- 24 June 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Nonequilibrium polysome dynamics promote chromosome segregation and its coupling to cell growth in Escherichia coli
Chromosome segregation is essential for cellular proliferation. Unlike eukaryotes, bacteria lack cytoskeleton-based machinery to segregate their chromosomal DNA (nucleoid). The bacterial ParABS system segregates the duplicated chromosomal regions near the origin of replication. However, this function does not explain how bacterial cells partition the rest (bulk) of the chromosomal material. Furthermore, some bacteria, including Escherichia coli, lack a ParABS system. Yet, E. coli faithfully segregates nucleoids across various growth rates. Here, we provide theoretical and experimental evidence that polysome production during chromosomal gene expression helps compact, split, segregate, and position nucleoids in E. coli through nonequilibrium dynamics that depend on polysome synthesis, degradation (through mRNA decay), and exclusion from the DNA meshwork. These dynamics inherently couple chromosome segregation to biomass growth across nutritional conditions. Halting chromosomal gene expression and thus polysome production immediately stops sister nucleoid migration, while ensuing polysome depletion gradually reverses nucleoid segregation. Redirecting gene expression away from the chromosome and toward plasmids causes ectopic polysome accumulations that are sufficient to drive aberrant nucleoid dynamics. Cell width enlargement experiments suggest that limiting the exchange of polysomes across DNA-free regions ensures nucleoid segregation along the cell length. Our findings suggest a self-organizing mechanism for coupling nucleoid compaction and segregation to cell growth without the apparent requirement of regulatory molecules.
- 24 June 2025
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 14
Open AccessNonequilibrium polysome dynamics promote chromosome segregation and its coupling to cell growth in Escherichia coli
Chromosome segregation is essential for cellular proliferation. Unlike eukaryotes, bacteria lack cytoskeleton-based machinery to segregate their chromosomal DNA (nucleoid). The bacterial ParABS system segregates the duplicated chromosomal regions near the origin of replication. However, this function does not explain how bacterial cells partition the rest (bulk) of the chromosomal material. Furthermore, some bacteria, including Escherichia coli, lack a ParABS system. Yet, E. coli faithfully segregates nucleoids across various growth rates. Here, we provide theoretical and experimental evidence that polysome production during chromosomal gene expression helps compact, split, segregate, and position nucleoids in E. coli through nonequilibrium dynamics that depend on polysome synthesis, degradation (through mRNA decay), and exclusion from the DNA meshwork. These dynamics inherently couple chromosome segregation to biomass growth across nutritional conditions. Halting chromosomal gene expression and thus polysome production immediately stops sister nucleoid migration, while ensuing polysome depletion gradually reverses nucleoid segregation. Redirecting gene expression away from the chromosome and toward plasmids causes ectopic polysome accumulations that are sufficient to drive aberrant nucleoid dynamics. Cell width enlargement experiments suggest that limiting the exchange of polysomes across DNA-free regions ensures nucleoid segregation along the cell length. Our findings suggest a self-organizing mechanism for coupling nucleoid compaction and segregation to cell growth without the apparent requirement of regulatory molecules.
- 19 May 2025
- Published by American Physical Society (APS) in Physical Review E
- Vol. 111 (5) , 054411
Exact minimax entropy models of large-scale neuronal activity
In the brain, fine-scale correlations combine to produce macroscopic patterns of activity. However, as experiments record from larger and larger populations, we approach a fundamental bottleneck: the number of correlations one would like to include in a model grows larger than the available data. In this undersampled regime, one must focus on a sparse subset of correlations; the optimal choice contains the maximum information about patterns of activity or, equivalently, minimizes the entropy of the inferred maximum entropy model. Applying this "minimax entropy" principle is generally intractable, but here we present an exact and scalable solution for pairwise correlations that combine to form a tree (a network without loops). Applying our method to over 1000 neurons in the mouse hippocampus, we find that the optimal tree of correlations reduces our uncertainty about the population activity by 14% (over 50 times more than a random tree). Despite containing only 0.1% of all pairwise correlations, this minimax entropy model accurately predicts the observed large-scale synchrony in neural activity and becomes even more accurate as the population grows. The inferred Ising model is almost entirely ferromagnetic (with positive interactions) and exhibits signatures of thermodynamic criticality. Together, these results suggest that a large amount of information may be compressed into a small number of correlations between neurons, and provide the tools for identifying the most important correlations in other complex living systems.
- 19 May 2025
- Published by American Physical Society (APS) in Physical Review E
- Vol. 111 (5) , 054411
Exact minimax entropy models of large-scale neuronal activity
In the brain, fine-scale correlations combine to produce macroscopic patterns of activity. However, as experiments record from larger and larger populations, we approach a fundamental bottleneck: the number of correlations one would like to include in a model grows larger than the available data. In this undersampled regime, one must focus on a sparse subset of correlations; the optimal choice contains the maximum information about patterns of activity or, equivalently, minimizes the entropy of the inferred maximum entropy model. Applying this "minimax entropy" principle is generally intractable, but here we present an exact and scalable solution for pairwise correlations that combine to form a tree (a network without loops). Applying our method to over 1000 neurons in the mouse hippocampus, we find that the optimal tree of correlations reduces our uncertainty about the population activity by 14% (over 50 times more than a random tree). Despite containing only 0.1% of all pairwise correlations, this minimax entropy model accurately predicts the observed large-scale synchrony in neural activity and becomes even more accurate as the population grows. The inferred Ising model is almost entirely ferromagnetic (with positive interactions) and exhibits signatures of thermodynamic criticality. Together, these results suggest that a large amount of information may be compressed into a small number of correlations between neurons, and provide the tools for identifying the most important correlations in other complex living systems.