@article{10166, abstract = {While sexual reproduction is widespread among many taxa, asexual lineages have repeatedly evolved from sexual ancestors. Despite extensive research on the evolution of sex, it is still unclear whether this switch represents a major transition requiring major molecular reorganization, and how convergent the changes involved are. In this study, we investigated the phylogenetic relationship and patterns of gene expression of sexual and asexual lineages of Eurasian Artemia brine shrimp, to assess how gene expression patterns are affected by the transition to asexuality. We find only a few genes that are consistently associated with the evolution of asexuality, suggesting that this shift may not require an extensive overhauling of the meiotic machinery. While genes with sex-biased expression have high rates of expression divergence within Eurasian Artemia, neither female- nor male-biased genes appear to show unusual evolutionary patterns after sexuality is lost, contrary to theoretical expectations.}, author = {Huylmans, Ann K and Macon, Ariana and Hontoria, Francisco and Vicoso, Beatriz}, issn = {1471-2954}, journal = {Proceedings of the Royal Society B: Biological Sciences}, keywords = {asexual reproduction, parthenogenesis, sex-biased genes, sexual conflict, automixis, crustaceans}, number = {1959}, publisher = {The Royal Society}, title = {{Transitions to asexuality and evolution of gene expression in Artemia brine shrimp}}, doi = {10.1098/rspb.2021.1720}, volume = {288}, year = {2021}, } @article{10167, abstract = {Schistosomes, the human parasites responsible for snail fever, are female-heterogametic. Different parts of their ZW sex chromosomes have stopped recombining in distinct lineages, creating “evolutionary strata” of various ages. Although the Z-chromosome is well characterized at the genomic and molecular level, the W-chromosome has remained largely unstudied from an evolutionary perspective, as only a few W-linked genes have been detected outside of the model species Schistosoma mansoni. Here, we characterize the gene content and evolution of the W-chromosomes of S. mansoni and of the divergent species S. japonicum. We use a combined RNA/DNA k-mer based pipeline to assemble around 100 candidate W-specific transcripts in each of the species. About half of them map to known protein coding genes, the majority homologous to S. mansoni Z-linked genes. We perform an extended analysis of the evolutionary strata present in the two species (including characterizing a previously undetected young stratum in S. japonicum) to infer patterns of sequence and expression evolution of W-linked genes at different time points after recombination was lost. W-linked genes show evidence of degeneration, including high rates of protein evolution and reduced expression. Most are found in young lineage-specific strata, with only a few high expression ancestral W-genes remaining, consistent with the progressive erosion of nonrecombining regions. Among these, the splicing factor u2af2 stands out as a promising candidate for primary sex determination, opening new avenues for understanding the molecular basis of the reproductive biology of this group.}, author = {Elkrewi, Marwan N and Moldovan, Mikhail A. and Picard, Marion A L and Vicoso, Beatriz}, issn = {1537-1719}, journal = {Molecular Biology and Evolution}, keywords = {sex chromosomes, evolutionary strata, W-linked gene, sex determining gene, schistosome parasites}, publisher = {Oxford University Press }, title = {{Schistosome W-Linked genes inform temporal dynamics of sex chromosome evolution and suggest candidate for sex determination}}, doi = {10.1093/molbev/msab178}, year = {2021}, } @article{10290, abstract = {A precise quantitative description of the ultrastructural characteristics underlying biological mechanisms is often key to their understanding. This is particularly true for dynamic extra- and intracellular filamentous assemblies, playing a role in cell motility, cell integrity, cytokinesis, tissue formation and maintenance. For example, genetic manipulation or modulation of actin regulatory proteins frequently manifests in changes of the morphology, dynamics, and ultrastructural architecture of actin filament-rich cell peripheral structures, such as lamellipodia or filopodia. However, the observed ultrastructural effects often remain subtle and require sufficiently large datasets for appropriate quantitative analysis. The acquisition of such large datasets has been enabled by recent advances in high-throughput cryo-electron tomography (cryo-ET) methods. This also necessitates the development of complementary approaches to maximize the extraction of relevant biological information. We have developed a computational toolbox for the semi-automatic quantification of segmented and vectorized filamentous networks from pre-processed cryo-electron tomograms, facilitating the analysis and cross-comparison of multiple experimental conditions. GUI-based components simplify the processing of data and allow users to obtain a large number of ultrastructural parameters describing filamentous assemblies. We demonstrate the feasibility of this workflow by analyzing cryo-ET data of untreated and chemically perturbed branched actin filament networks and that of parallel actin filament arrays. In principle, the computational toolbox presented here is applicable for data analysis comprising any type of filaments in regular (i.e. parallel) or random arrangement. We show that it can ease the identification of key differences between experimental groups and facilitate the in-depth analysis of ultrastructural data in a time-efficient manner.}, author = {Dimchev, Georgi A and Amiri, Behnam and Fäßler, Florian and Falcke, Martin and Schur, Florian KM}, issn = {1047-8477}, journal = {Journal of Structural Biology}, keywords = {Structural Biology}, number = {4}, publisher = {Elsevier }, title = {{Computational toolbox for ultrastructural quantitative analysis of filament networks in cryo-ET data}}, doi = {10.1016/j.jsb.2021.107808}, volume = {213}, year = {2021}, } @article{10535, abstract = {Realistic models of biological processes typically involve interacting components on multiple scales, driven by changing environment and inherent stochasticity. Such models are often analytically and numerically intractable. We revisit a dynamic maximum entropy method that combines a static maximum entropy with a quasi-stationary approximation. This allows us to reduce stochastic non-equilibrium dynamics expressed by the Fokker-Planck equation to a simpler low-dimensional deterministic dynamics, without the need to track microscopic details. Although the method has been previously applied to a few (rather complicated) applications in population genetics, our main goal here is to explain and to better understand how the method works. We demonstrate the usefulness of the method for two widely studied stochastic problems, highlighting its accuracy in capturing important macroscopic quantities even in rapidly changing non-stationary conditions. For the Ornstein-Uhlenbeck process, the method recovers the exact dynamics whilst for a stochastic island model with migration from other habitats, the approximation retains high macroscopic accuracy under a wide range of scenarios in a dynamic environment.}, author = {Bod'ová, Katarína and Szep, Eniko and Barton, Nicholas H}, issn = {1553-7358}, journal = {PLoS Computational Biology}, number = {12}, publisher = {Public Library of Science}, title = {{Dynamic maximum entropy provides accurate approximation of structured population dynamics}}, doi = {10.1371/journal.pcbi.1009661}, volume = {17}, year = {2021}, } @phdthesis{9418, abstract = {Deep learning is best known for its empirical success across a wide range of applications spanning computer vision, natural language processing and speech. Of equal significance, though perhaps less known, are its ramifications for learning theory: deep networks have been observed to perform surprisingly well in the high-capacity regime, aka the overfitting or underspecified regime. Classically, this regime on the far right of the bias-variance curve is associated with poor generalisation; however, recent experiments with deep networks challenge this view. This thesis is devoted to investigating various aspects of underspecification in deep learning. First, we argue that deep learning models are underspecified on two levels: a) any given training dataset can be fit by many different functions, and b) any given function can be expressed by many different parameter configurations. We refer to the second kind of underspecification as parameterisation redundancy and we precisely characterise its extent. Second, we characterise the implicit criteria (the inductive bias) that guide learning in the underspecified regime. Specifically, we consider a nonlinear but tractable classification setting, and show that given the choice, neural networks learn classifiers with a large margin. Third, we consider learning scenarios where the inductive bias is not by itself sufficient to deal with underspecification. We then study different ways of ‘tightening the specification’: i) In the setting of representation learning with variational autoencoders, we propose a hand- crafted regulariser based on mutual information. ii) In the setting of binary classification, we consider soft-label (real-valued) supervision. We derive a generalisation bound for linear networks supervised in this way and verify that soft labels facilitate fast learning. Finally, we explore an application of soft-label supervision to the training of multi-exit models.}, author = {Bui Thi Mai, Phuong}, issn = {2663-337X}, pages = {125}, publisher = {Institute of Science and Technology Austria}, title = {{Underspecification in deep learning}}, doi = {10.15479/AT:ISTA:9418}, year = {2021}, } @article{9431, abstract = {Inositol hexakisphosphate (IP6) is an assembly cofactor for HIV-1. We report here that IP6 is also used for assembly of Rous sarcoma virus (RSV), a retrovirus from a different genus. IP6 is ~100-fold more potent at promoting RSV mature capsid protein (CA) assembly than observed for HIV-1 and removal of IP6 in cells reduces infectivity by 100-fold. Here, visualized by cryo-electron tomography and subtomogram averaging, mature capsid-like particles show an IP6-like density in the CA hexamer, coordinated by rings of six lysines and six arginines. Phosphate and IP6 have opposing effects on CA in vitro assembly, inducing formation of T = 1 icosahedrons and tubes, respectively, implying that phosphate promotes pentamer and IP6 hexamer formation. Subtomogram averaging and classification optimized for analysis of pleomorphic retrovirus particles reveal that the heterogeneity of mature RSV CA polyhedrons results from an unexpected, intrinsic CA hexamer flexibility. In contrast, the CA pentamer forms rigid units organizing the local architecture. These different features of hexamers and pentamers determine the structural mechanism to form CA polyhedrons of variable shape in mature RSV particles.}, author = {Obr, Martin and Ricana, Clifton L. and Nikulin, Nadia and Feathers, Jon-Philip R. and Klanschnig, Marco and Thader, Andreas and Johnson, Marc C. and Vogt, Volker M. and Schur, Florian KM and Dick, Robert A.}, issn = {2041-1723}, journal = {Nature Communications}, keywords = {General Biochemistry, Genetics and Molecular Biology, General Physics and Astronomy, General Chemistry}, number = {1}, publisher = {Nature Research}, title = {{Structure of the mature Rous sarcoma virus lattice reveals a role for IP6 in the formation of the capsid hexamer}}, doi = {10.1038/s41467-021-23506-0}, volume = {12}, year = {2021}, } @article{9558, abstract = {We show that turbulent dynamics that arise in simulations of the three-dimensional Navier--Stokes equations in a triply-periodic domain under sinusoidal forcing can be described as transient visits to the neighborhoods of unstable time-periodic solutions. Based on this description, we reduce the original system with more than 10^5 degrees of freedom to a 17-node Markov chain where each node corresponds to the neighborhood of a periodic orbit. The model accurately reproduces long-term averages of the system's observables as weighted sums over the periodic orbits. }, author = {Yalniz, Gökhan and Hof, Björn and Budanur, Nazmi B}, issn = {1079-7114}, journal = {Physical Review Letters}, number = {24}, publisher = {American Physical Society}, title = {{Coarse graining the state space of a turbulent flow using periodic orbits}}, doi = {10.1103/PhysRevLett.126.244502}, volume = {126}, year = {2021}, } @article{9818, abstract = {Triangle mesh-based simulations are able to produce satisfying animations of knitted and woven cloth; however, they lack the rich geometric detail of yarn-level simulations. Naive texturing approaches do not consider yarn-level physics, while full yarn-level simulations may become prohibitively expensive for large garments. We propose a method to animate yarn-level cloth geometry on top of an underlying deforming mesh in a mechanics-aware fashion. Using triangle strains to interpolate precomputed yarn geometry, we are able to reproduce effects such as knit loops tightening under stretching. In combination with precomputed mesh animation or real-time mesh simulation, our method is able to animate yarn-level cloth in real-time at large scales.}, author = {Sperl, Georg and Narain, Rahul and Wojtan, Christopher J}, issn = {15577368}, journal = {ACM Transactions on Graphics}, number = {4}, publisher = {Association for Computing Machinery}, title = {{Mechanics-aware deformation of yarn pattern geometry}}, doi = {10.1145/3450626.3459816}, volume = {40}, year = {2021}, } @inproceedings{11458, abstract = {The increasing computational requirements of deep neural networks (DNNs) have led to significant interest in obtaining DNN models that are sparse, yet accurate. Recent work has investigated the even harder case of sparse training, where the DNN weights are, for as much as possible, already sparse to reduce computational costs during training. Existing sparse training methods are often empirical and can have lower accuracy relative to the dense baseline. In this paper, we present a general approach called Alternating Compressed/DeCompressed (AC/DC) training of DNNs, demonstrate convergence for a variant of the algorithm, and show that AC/DC outperforms existing sparse training methods in accuracy at similar computational budgets; at high sparsity levels, AC/DC even outperforms existing methods that rely on accurate pre-trained dense models. An important property of AC/DC is that it allows co-training of dense and sparse models, yielding accurate sparse–dense model pairs at the end of the training process. This is useful in practice, where compressed variants may be desirable for deployment in resource-constrained settings without re-doing the entire training flow, and also provides us with insights into the accuracy gap between dense and compressed models. The code is available at: https://github.com/IST-DASLab/ACDC.}, author = {Peste, Elena-Alexandra and Iofinova, Eugenia B and Vladu, Adrian and Alistarh, Dan-Adrian}, booktitle = {35th Conference on Neural Information Processing Systems}, isbn = {9781713845393}, issn = {1049-5258}, location = {Virtual, Online}, pages = {8557--8570}, publisher = {Curran Associates}, title = {{AC/DC: Alternating Compressed/DeCompressed training of deep neural networks}}, volume = {34}, year = {2021}, } @article{8261, abstract = {Dentate gyrus granule cells (GCs) connect the entorhinal cortex to the hippocampal CA3 region, but how they process spatial information remains enigmatic. To examine the role of GCs in spatial coding, we measured excitatory postsynaptic potentials (EPSPs) and action potentials (APs) in head-fixed mice running on a linear belt. Intracellular recording from morphologically identified GCs revealed that most cells were active, but activity level varied over a wide range. Whereas only ∼5% of GCs showed spatially tuned spiking, ∼50% received spatially tuned input. Thus, the GC population broadly encodes spatial information, but only a subset relays this information to the CA3 network. Fourier analysis indicated that GCs received conjunctive place-grid-like synaptic input, suggesting code conversion in single neurons. GC firing was correlated with dendritic complexity and intrinsic excitability, but not extrinsic excitatory input or dendritic cable properties. Thus, functional maturation may control input-output transformation and spatial code conversion.}, author = {Zhang, Xiaomin and Schlögl, Alois and Jonas, Peter M}, issn = {0896-6273}, journal = {Neuron}, number = {6}, pages = {1212--1225}, publisher = {Elsevier}, title = {{Selective routing of spatial information flow from input to output in hippocampal granule cells}}, doi = {10.1016/j.neuron.2020.07.006}, volume = {107}, year = {2020}, } @phdthesis{8353, abstract = {Mrp (Multi resistance and pH adaptation) are broadly distributed secondary active antiporters that catalyze the transport of monovalent ions such as sodium and potassium outside of the cell coupled to the inward translocation of protons. Mrp antiporters are unique in a way that they are composed of seven subunits (MrpABCDEFG) encoded in a single operon, whereas other antiporters catalyzing the same reaction are mostly encoded by a single gene. Mrp exchangers are crucial for intracellular pH homeostasis and Na+ efflux, essential mechanisms for H+ uptake under alkaline environments and for reduction of the intracellular concentration of toxic cations. Mrp displays no homology to any other monovalent Na+(K+)/H+ antiporters but Mrp subunits have primary sequence similarity to essential redox-driven proton pumps, such as respiratory complex I and membrane-bound hydrogenases. This similarity reinforces the hypothesis that these present day redox-driven proton pumps are descended from the Mrp antiporter. The Mrp structure serves as a model to understand the yet obscure coupling mechanism between ion or electron transfer and proton translocation in this large group of proteins. In the thesis, I am presenting the purification, biochemical analysis, cryo-EM analysis and molecular structure of the Mrp complex from Anoxybacillus flavithermus solved by cryo-EM at 3.0 Å resolution. Numerous conditions were screened to purify Mrp to high homogeneity and to obtain an appropriate distribution of single particles on cryo-EM grids covered with a continuous layer of ultrathin carbon. A preferred particle orientation problem was solved by performing a tilted data collection. The activity assays showed the specific pH-dependent profile of secondary active antiporters. The molecular structure shows that Mrp is a dimer of seven-subunit protomers with 50 trans-membrane helices each. The dimer interface is built by many short and tilted transmembrane helices, probably causing a thinning of the bacterial membrane. The surface charge distribution shows an extraordinary asymmetry within each monomer, revealing presumable proton and sodium translocation pathways. The two largest and homologous Mrp subunits MrpA and MrpD probably translocate one proton each into the cell. The sodium ion is likely being translocated in the opposite direction within the small subunits along a ladder of charged and conserved residues. Based on the structure, we propose a mechanism were the antiport activity is accomplished via electrostatic interactions between the charged cations and key charged residues. The flexible key TM helices coordinate these electrostatic interactions, while the membrane thinning between the monomers enables the translocation of sodium across the charged membrane. The entire family of redox-driven proton pumps is likely to perform their mechanism in a likewise manner.}, author = {Steiner, Julia}, issn = {2663-337X}, pages = {191}, publisher = {Institute of Science and Technology Austria}, title = {{Biochemical and structural investigation of the Mrp antiporter, an ancestor of complex I}}, doi = {10.15479/AT:ISTA:8353}, year = {2020}, } @phdthesis{8366, abstract = {Fabrication of curved shells plays an important role in modern design, industry, and science. Among their remarkable properties are, for example, aesthetics of organic shapes, ability to evenly distribute loads, or efficient flow separation. They find applications across vast length scales ranging from sky-scraper architecture to microscopic devices. But, at the same time, the design of curved shells and their manufacturing process pose a variety of challenges. In this thesis, they are addressed from several perspectives. In particular, this thesis presents approaches based on the transformation of initially flat sheets into the target curved surfaces. This involves problems of interactive design of shells with nontrivial mechanical constraints, inverse design of complex structural materials, and data-driven modeling of delicate and time-dependent physical properties. At the same time, two newly-developed self-morphing mechanisms targeting flat-to-curved transformation are presented. In architecture, doubly curved surfaces can be realized as cold bent glass panelizations. Originally flat glass panels are bent into frames and remain stressed. This is a cost-efficient fabrication approach compared to hot bending, when glass panels are shaped plastically. However such constructions are prone to breaking during bending, and it is highly nontrivial to navigate the design space, keeping the panels fabricable and aesthetically pleasing at the same time. We introduce an interactive design system for cold bent glass façades, while previously even offline optimization for such scenarios has not been sufficiently developed. Our method is based on a deep learning approach providing quick and high precision estimation of glass panel shape and stress while handling the shape multimodality. Fabrication of smaller objects of scales below 1 m, can also greatly benefit from shaping originally flat sheets. In this respect, we designed new self-morphing shell mechanisms transforming from an initial flat state to a doubly curved state with high precision and detail. Our so-called CurveUps demonstrate the encodement of the geometric information into the shell. Furthermore, we explored the frontiers of programmable materials and showed how temporal information can additionally be encoded into a flat shell. This allows prescribing deformation sequences for doubly curved surfaces and, thus, facilitates self-collision avoidance enabling complex shapes and functionalities otherwise impossible. Both of these methods include inverse design tools keeping the user in the design loop.}, author = {Guseinov, Ruslan}, isbn = {978-3-99078-010-7}, issn = {2663-337X}, keywords = {computer-aided design, shape modeling, self-morphing, mechanical engineering}, pages = {118}, publisher = {Institute of Science and Technology Austria}, title = {{Computational design of curved thin shells: From glass façades to programmable matter}}, doi = {10.15479/AT:ISTA:8366}, year = {2020}, } @article{8384, abstract = {Previous research on animations of soap bubbles, films, and foams largely focuses on the motion and geometric shape of the bubble surface. These works neglect the evolution of the bubble’s thickness, which is normally responsible for visual phenomena like surface vortices, Newton’s interference patterns, capillary waves, and deformation-dependent rupturing of films in a foam. In this paper, we model these natural phenomena by introducing the film thickness as a reduced degree of freedom in the Navier-Stokes equations and deriving their equations of motion. We discretize the equations on a nonmanifold triangle mesh surface and couple it to an existing bubble solver. In doing so, we also introduce an incompressible fluid solver for 2.5D films and a novel advection algorithm for convecting fields across non-manifold surface junctions. Our simulations enhance state-of-the-art bubble solvers with additional effects caused by convection, rippling, draining, and evaporation of the thin film.}, author = {Ishida, Sadashige and Synak, Peter and Narita, Fumiya and Hachisuka, Toshiya and Wojtan, Christopher J}, issn = {15577368}, journal = {ACM Transactions on Graphics}, number = {4}, publisher = {Association for Computing Machinery}, title = {{A model for soap film dynamics with evolving thickness}}, doi = {10.1145/3386569.3392405}, volume = {39}, year = {2020}, } @article{8385, abstract = {We present a method for animating yarn-level cloth effects using a thin-shell solver. We accomplish this through numerical homogenization: we first use a large number of yarn-level simulations to build a model of the potential energy density of the cloth, and then use this energy density function to compute forces in a thin shell simulator. We model several yarn-based materials, including both woven and knitted fabrics. Our model faithfully reproduces expected effects like the stiffness of woven fabrics, and the highly deformable nature and anisotropy of knitted fabrics. Our approach does not require any real-world experiments nor measurements; because the method is based entirely on simulations, it can generate entirely new material models quickly, without the need for testing apparatuses or human intervention. We provide data-driven models of several woven and knitted fabrics, which can be used for efficient simulation with an off-the-shelf cloth solver.}, author = {Sperl, Georg and Narain, Rahul and Wojtan, Christopher J}, issn = {15577368}, journal = {ACM Transactions on Graphics}, number = {4}, publisher = {Association for Computing Machinery}, title = {{Homogenized yarn-level cloth}}, doi = {10.1145/3386569.3392412}, volume = {39}, year = {2020}, } @phdthesis{8390, abstract = {Deep neural networks have established a new standard for data-dependent feature extraction pipelines in the Computer Vision literature. Despite their remarkable performance in the standard supervised learning scenario, i.e. when models are trained with labeled data and tested on samples that follow a similar distribution, neural networks have been shown to struggle with more advanced generalization abilities, such as transferring knowledge across visually different domains, or generalizing to new unseen combinations of known concepts. In this thesis we argue that, in contrast to the usual black-box behavior of neural networks, leveraging more structured internal representations is a promising direction for tackling such problems. In particular, we focus on two forms of structure. First, we tackle modularity: We show that (i) compositional architectures are a natural tool for modeling reasoning tasks, in that they efficiently capture their combinatorial nature, which is key for generalizing beyond the compositions seen during training. We investigate how to to learn such models, both formally and experimentally, for the task of abstract visual reasoning. Then, we show that (ii) in some settings, modularity allows us to efficiently break down complex tasks into smaller, easier, modules, thereby improving computational efficiency; We study this behavior in the context of generative models for colorization, as well as for small objects detection. Secondly, we investigate the inherently layered structure of representations learned by neural networks, and analyze its role in the context of transfer learning and domain adaptation across visually dissimilar domains. }, author = {Royer, Amélie}, isbn = {978-3-99078-007-7}, issn = {2663-337X}, pages = {197}, publisher = {Institute of Science and Technology Austria}, title = {{Leveraging structure in Computer Vision tasks for flexible Deep Learning models}}, doi = {10.15479/AT:ISTA:8390}, year = {2020}, } @article{8535, abstract = {We propose a method to enhance the visual detail of a water surface simulation. Our method works as a post-processing step which takes a simulation as input and increases its apparent resolution by simulating many detailed Lagrangian water waves on top of it. We extend linear water wave theory to work in non-planar domains which deform over time, and we discretize the theory using Lagrangian wave packets attached to spline curves. The method is numerically stable and trivially parallelizable, and it produces high frequency ripples with dispersive wave-like behaviors customized to the underlying fluid simulation.}, author = {Skrivan, Tomas and Soderstrom, Andreas and Johansson, John and Sprenger, Christoph and Museth, Ken and Wojtan, Christopher J}, issn = {15577368}, journal = {ACM Transactions on Graphics}, number = {4}, publisher = {Association for Computing Machinery}, title = {{Wave curves: Simulating Lagrangian water waves on dynamically deforming surfaces}}, doi = {10.1145/3386569.3392466}, volume = {39}, year = {2020}, } @article{8562, abstract = {Cold bent glass is a promising and cost-efficient method for realizing doubly curved glass facades. They are produced by attaching planar glass sheets to curved frames and require keeping the occurring stress within safe limits. However, it is very challenging to navigate the design space of cold bent glass panels due to the fragility of the material, which impedes the form-finding for practically feasible and aesthetically pleasing cold bent glass facades. We propose an interactive, data-driven approach for designing cold bent glass facades that can be seamlessly integrated into a typical architectural design pipeline. Our method allows non-expert users to interactively edit a parametric surface while providing real-time feedback on the deformed shape and maximum stress of cold bent glass panels. Designs are automatically refined to minimize several fairness criteria while maximal stresses are kept within glass limits. We achieve interactive frame rates by using a differentiable Mixture Density Network trained from more than a million simulations. Given a curved boundary, our regression model is capable of handling multistable configurations and accurately predicting the equilibrium shape of the panel and its corresponding maximal stress. We show predictions are highly accurate and validate our results with a physical realization of a cold bent glass surface.}, author = {Gavriil, Konstantinos and Guseinov, Ruslan and Perez Rodriguez, Jesus and Pellis, Davide and Henderson, Paul M and Rist, Florian and Pottmann, Helmut and Bickel, Bernd}, issn = {1557-7368}, journal = {ACM Transactions on Graphics}, number = {6}, publisher = {Association for Computing Machinery}, title = {{Computational design of cold bent glass façades}}, doi = {10.1145/3414685.3417843}, volume = {39}, year = {2020}, } @article{8581, abstract = {The majority of adenosine triphosphate (ATP) powering cellular processes in eukaryotes is produced by the mitochondrial F1Fo ATP synthase. Here, we present the atomic models of the membrane Fo domain and the entire mammalian (ovine) F1Fo, determined by cryo-electron microscopy. Subunits in the membrane domain are arranged in the ‘proton translocation cluster’ attached to the c-ring and a more distant ‘hook apparatus’ holding subunit e. Unexpectedly, this subunit is anchored to a lipid ‘plug’ capping the c-ring. We present a detailed proton translocation pathway in mammalian Fo and key inter-monomer contacts in F1Fo multimers. Cryo-EM maps of F1Fo exposed to calcium reveal a retracted subunit e and a disassembled c-ring, suggesting permeability transition pore opening. We propose a model for the permeability transition pore opening, whereby subunit e pulls the lipid plug out of the c-ring. Our structure will allow the design of drugs for many emerging applications in medicine.}, author = {Pinke, Gergely and Zhou, Long and Sazanov, Leonid A}, issn = {15459985}, journal = {Nature Structural and Molecular Biology}, number = {11}, pages = {1077--1085}, publisher = {Springer Nature}, title = {{Cryo-EM structure of the entire mammalian F-type ATP synthase}}, doi = {10.1038/s41594-020-0503-8}, volume = {27}, year = {2020}, } @article{8586, abstract = {Cryo-electron microscopy (cryo-EM) of cellular specimens provides insights into biological processes and structures within a native context. However, a major challenge still lies in the efficient and reproducible preparation of adherent cells for subsequent cryo-EM analysis. This is due to the sensitivity of many cellular specimens to the varying seeding and culturing conditions required for EM experiments, the often limited amount of cellular material and also the fragility of EM grids and their substrate. Here, we present low-cost and reusable 3D printed grid holders, designed to improve specimen preparation when culturing challenging cellular samples directly on grids. The described grid holders increase cell culture reproducibility and throughput, and reduce the resources required for cell culturing. We show that grid holders can be integrated into various cryo-EM workflows, including micro-patterning approaches to control cell seeding on grids, and for generating samples for cryo-focused ion beam milling and cryo-electron tomography experiments. Their adaptable design allows for the generation of specialized grid holders customized to a large variety of applications.}, author = {Fäßler, Florian and Zens, Bettina and Hauschild, Robert and Schur, Florian KM}, issn = {1047-8477}, journal = {Journal of Structural Biology}, keywords = {electron microscopy, cryo-EM, EM sample preparation, 3D printing, cell culture}, number = {3}, publisher = {Elsevier}, title = {{3D printed cell culture grid holders for improved cellular specimen preparation in cryo-electron microscopy}}, doi = {10.1016/j.jsb.2020.107633}, volume = {212}, year = {2020}, } @inproceedings{8724, abstract = {We study the problem of learning from multiple untrusted data sources, a scenario of increasing practical relevance given the recent emergence of crowdsourcing and collaborative learning paradigms. Specifically, we analyze the situation in which a learning system obtains datasets from multiple sources, some of which might be biased or even adversarially perturbed. It is known that in the single-source case, an adversary with the power to corrupt a fixed fraction of the training data can prevent PAC-learnability, that is, even in the limit of infinitely much training data, no learning system can approach the optimal test error. In this work we show that, surprisingly, the same is not true in the multi-source setting, where the adversary can arbitrarily corrupt a fixed fraction of the data sources. Our main results are a generalization bound that provides finite-sample guarantees for this learning setting, as well as corresponding lower bounds. Besides establishing PAC-learnability our results also show that in a cooperative learning setting sharing data with other parties has provable benefits, even if some participants are malicious. }, author = {Konstantinov, Nikola H and Frantar, Elias and Alistarh, Dan-Adrian and Lampert, Christoph}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, issn = {2640-3498}, location = {Online}, pages = {5416--5425}, publisher = {ML Research Press}, title = {{On the sample complexity of adversarial multi-source PAC learning}}, volume = {119}, year = {2020}, }