Earth Imagery Segmentation on Terrain Surface with Limited Training Labels: A Semi-supervised Approach based on Physics-Guided Graph Co-Training

Given earth imagery with spectral features on a terrain surface, this paper studies surface segmentation based on both explanatory features and surface topology. The problem is important in many spatial and spatiotemporal applications such as flood extent mapping in hydrology. The problem is uniquely challenging for several reasons: first, the size of earth imagery on a terrain surface is often much larger than the input of popular deep convolutional neural networks; second, there exists topological structure dependency between pixel classes on the surface, and such dependency can follow an unknown and non-linear distribution; third, there are often limited training labels. Existing methods for earth imagery segmentation often divide the imagery into patches and consider the elevation as an additional feature channel. These methods do not fully incorporate the spatial topological structural constraint within and across surface patches and thus often show poor results, especially when training labels are limited. Existing methods on semi-supervised and unsupervised learning for earth imagery often focus on learning representation without explicitly incorporating surface topology. In contrast, we propose a novel framework that explicitly models the topological skeleton of a terrain surface with a contour tree from computational topology, which is guided by the physical constraint (e.g., water flow direction on terrains). Our framework consists of two neural networks: a convolutional neural network (CNN) to learn spatial contextual features on a 2D image grid, and a graph neural network (GNN) to learn the statistical distribution of physics-guided spatial topological dependency on the contour tree. The two models are co-trained via variational EM. Evaluations on the real-world flood mapping datasets show that the proposed models outperform baseline methods in classification accuracy, especially when training labels are limited.

Crystal structure and Hirshfeld surfaces analysis of Heterocyclic-and circulenes

The crystal structure of the new diazatrioxa[9]circulene and tetrahydro-diazatetraoxa[10]circulene which represent the first synthesized representatives of “higher” hetero[n]circulenes with n>8, was analyzed in details. Hirshfeld surface analyses, the dnorm surfaces and two-dimensional fingerprint plots were used to verify the contributions of the different intermolecular interactions within the crystal structure of diazatrioxa[9]circulene and tetrahydro-diazatetraoxa[10]circulene. The Hirshfeld surface analysis of the crystal structure clarifies that the most important contribution for crystal packing is from H∙∙∙H and C∙∙∙H intermolecular interactions for both circulenes. The shape-index surface shows that in the case of diazatrioxa[9]circulene two sides of the molecules are involved with the same contacts in neighbouring molecules and curvedness plots show flat surface patches that are characteristic of planar stacking. Such face-to-face structural organization provides the main charge transfer pathway in [9]circulene. In the case of [10]circulene, the area involved in the same contacts is much less, however, two types of intermolecular packing modes can form such flat surface patches at curvedness plots which is useful for more efficient charge transfer.

Fast Quadratic Shape From Shading With Unknown Light Direction

An algorithm is proposed that extracts 3D shape from shading information in a digital image. The algorithm assumes that there is only a single source of light producing the image, that the surface of the shape giving rise to the image is Lambertian (matte) and that its shape can be locally approximated by a quadratic function. Previous work shows that under these assumptions, robust shape from shading is possible, though slow for large images because a non-linear optimization method is applied in order to estimate local quadratic surface patches from image intensities. The work presented here shows that local quadratic surface patch estimates can be computed, without prior knowledge of the light source direction, via a linear least squares optimization, thus greatly improving the algebraic complexity and run-time of the existing algorithms.

Fast Quadratic Shape From Shading With Unknown Light Direction

An algorithm is proposed that extracts 3D shape from shading information in a digital image. The algorithm assumes that there is only a single source of light producing the image, that the surface of the shape giving rise to the image is Lambertian (matte) and that its shape can be locally approximated by a quadratic function. Previous work shows that under these assumptions, robust shape from shading is possible, though slow for large images because a non-linear optimization method is applied in order to estimate local quadratic surface patches from image intensities. The work presented here shows that local quadratic surface patch estimates can be computed, without prior knowledge of the light source direction, via a linear least squares optimization, thus greatly improving the algebraic complexity and run-time of this existing algorithms.

Fast Quadratic Shape From Shading With Unknown Light Direction

An algorithm is proposed that extracts 3D shape from shading information in a digital image. The algorithm assumes that there is only a single source of light producing the image, that the surface of the shape giving rise to the image is Lambertian (matte) and that its shape can be locally approximated by a quadratic function. Previous work shows that under these assumptions, robust shape from shading is possible, though slow for large images because a non-linear optimization method is applied in order to estimate local quadratic surface patches from image intensities. The work presented here shows that local quadratic surface patch estimates can be computed, without prior knowledge of the light source direction, via a linear least squares optimization, thus greatly improving the algebraic complexity and run-time of this existing algorithms.

Computational Studies of the Structural Basis of Human RPS19 Mutations Associated With Diamond-Blackfan Anemia

Diamond-Blackfan Anemia (DBA) is an inherited rare disease characterized with severe pure red cell aplasia, and it is caused by the defective ribosome biogenesis stemming from the impairment of ribosomal proteins. Among all DBA-associated ribosomal proteins, RPS19 affects most patients and carries most DBA mutations. Revealing how these mutations lead to the impairment of RPS19 is highly demanded for understanding the pathogenesis of DBA, but a systematic study is currently lacking. In this work, based on the complex structure of human ribosome, we comprehensively studied the structural basis of DBA mutations of RPS19 by using computational methods. Main structure elements and five conserved surface patches involved in RPS19-18S rRNA interaction were identified. We further revealed that DBA mutations would destabilize RPS19 through disrupting the hydrophobic core or breaking the helix, or perturb the RPS19-18S rRNA interaction through destroying hydrogen bonds, introducing steric hindrance effect, or altering surface electrostatic property at the interface. Moreover, we trained a machine-learning model to predict the pathogenicity of all possible RPS19 mutations. Our work has laid a foundation for revealing the pathogenesis of DBA from the structural perspective.

Multi-objective aerodynamic optimization of high-speed train heads based on the PDE parametric modeling

With the increasing running speed, the aerodynamic issues of high-speed trains are being raised and impact the running stability and energy efficiency. The optimization design of the head shape is significantly important in improving the aerodynamic performance of high-speed trains. Existing aerodynamic optimization methods are limited by the parametric modeling methods of train heads which are unable to accurately and completely parameterize both global shapes and local details. Due to this reason, they cannot optimize both global and local shapes of train heads. In order to tackle this problem, we propose a novel multi-objective aerodynamic optimization method of high-speed train heads based on the partial differential equation (PDE) parametric modeling. With this method, the half of a train head is parameterized with 17 PDE surface patches which describe global shapes and local details and keep the surface smooth. We take the aerodynamic drag and lift as optimization objectives; divide the optimization design process into two stages: global optimization and local optimization; and develop global and local optimization methods, respectively. In the first stage, the non-dominated sorting genetic algorithm (NSGA-II) is adopted to obtain the framework of the train head with an optimized global shape. In the second stage, Latin hypercube sampling (LHS) is applied in the local shape optimization of the PDE surface patches determined by the optimized framework to improve local details. The effectiveness of our proposed method is demonstrated by better aerodynamic performance achieved from the optimization solutions in global and local optimization stages in comparison with the original high-speed train head.

Basigin drives intracellular accumulation of l-lactate by harvesting protons and substrate anions

Transmembrane transport of l-lactate by members of the monocarboxylate transporter family, MCT, is vital in human physiology and a malignancy factor in cancer. Interaction with an accessory protein, typically basigin, is required to deliver the MCT to the plasma membrane. It is unknown whether basigin additionally exerts direct effects on the transmembrane l-lactate transport of MCT1. Here, we show that the presence of basigin leads to an intracellular accumulation of l-lactate 4.5-fold above the substrate/proton concentrations provided by the external buffer. Using basigin truncations we localized the effect to arise from the extracellular Ig-I domain. Identification of surface patches of condensed opposite electrostatic potential, and experimental analysis of charge-affecting Ig-I mutants indicated a bivalent harvesting antenna functionality for both, protons and substrate anions. From these data, and determinations of the cytosolic pH with a fluorescent probe, we conclude that the basigin Ig-I domain drives lactate uptake by locally increasing the proton and substrate concentration at the extracellular MCT entry site. The biophysical properties are physiologically relevant as cell growth on lactate media was strongly promoted in the presence of the Ig-I domain. Lack of the domain due to shedding, or misfolding due to breakage of a stabilizing disulfide bridge reversed the effect. Tumor progression according to classical or reverse Warburg effects depends on the transmembrane l-lactate distribution, and this study shows that the basigin Ig-I domain is a pivotal determinant.

Impaired Cytoskeletal and Membrane Biophysical Properties of Acanthocytes in Hypobetalipoproteinemia – A Case Study

Familial hypobetalipoproteinemia is a metabolic disorder mainly caused by mutations in the apolipoprotein B gene. In its homozygous form it can lead without treatment to severe ophthalmological and neurological manifestations. In contrast, the heterozygous form is generally asymptomatic but associated with a low risk of cardiovascular disease. Acanthocytes or thorny red blood cells (RBCs) are described for both forms of the disease. However, those morphological changes are poorly characterized and their potential consequences for RBC functionality are not understood. Thus, in the present study, we asked whether, to what extent and how acanthocytes from a patient with heterozygous familial hypobetalipoproteinemia could exhibit altered RBC functionality. Acanthocytes represented 50% of the total RBC population and contained mitoTracker-positive surface patches, indicating the presence of mitochondrial fragments. While RBC osmotic fragility, calcium content and ATP homeostasis were preserved, a slight decrease of RBC deformability combined with an increase of intracellular free reactive oxygen species were observed. The spectrin cytoskeleton was altered, showing a lower density and an enrichment in patches. At the membrane level, no obvious modification of the RBC membrane fatty acids nor of the cholesterol content were detected but the ceramide species were all increased. Membrane stiffness and curvature were also increased whereas transversal asymmetry was preserved. In contrast, lateral asymmetry was highly impaired showing: (i) increased abundance and decreased functionality of sphingomyelin-enriched domains; (ii) cholesterol enrichment in spicules; and (iii) ceramide enrichment in patches. We propose that oxidative stress induces cytoskeletal alterations, leading to increased membrane stiffness and curvature and impaired lipid lateral distribution in domains and spicules. In addition, ceramide- and spectrin-enriched patches could result from a RBC maturation defect. Altogether, the data indicate that acanthocytes are associated with cytoskeletal and membrane lipid lateral asymmetry alterations, while deformability is only mildly impaired. In addition, familial hypobetalipoproteinemia might also affect RBC precursors leading to disturbed RBC maturation. This study paves the way for the potential use of membrane biophysics and lipid vital imaging as new methods for diagnosis of RBC disorders.