Differentiable Programming of Isometric Tensor Networks

Differentiable programming is a new programming paradigm which enables large scale optimization through automatic calculation of gradients also known as auto-differentiation. This concept emerges from deep learning, and has also been generalized to tensor network optimizations. Here, we extend the differentiable programming to tensor networks with isometric constraints with applications to multiscale entanglement renormalization ansatz (MERA) and tensor network renormalization (TNR). By introducing several gradient-based optimization methods for the isometric tensor network and comparing with Evenbly-Vidal method, we show that auto-differentiation has a better performance for both stability and accuracy. We numerically tested our methods on 1D critical quantum Ising spin chain and 2D classical Ising model. We calculate the ground state energy for the 1D quantum model and internal energy for the classical model, and scaling dimensions of scaling operators and find they all agree with the theory well.

Conserved meiotic mechanisms in the cnidarian Clytia hemisphaerica revealed by Spo11 knockout

During meiosis, each duplicated chromosome pairs and recombines with its unique homolog to ensure the shuffling of genetic information across generations. Functional studies in classical model organisms have revealed a surprising diversity in the chronology and interdependency of the earliest meiotic steps such as chromosome movements, pairing, association via Synaptonemal Complex formation (synapsis), recombination and the formation of chiasmata. A key player is Spo11, an evolutionarily conserved topoisomerase-related transesterase that initiates meiotic recombination via the catalysis of programmed DNA double stranded breaks (DSBs). While DSBs are required for pairing and synapsis in budding yeast and mouse, alternative pathways are employed during female meiosis of the fruit fly and nematode Caenorhabditis elegans. Here, to provide a comparative perspective on meiotic regulation from a distinct animal clade, we chart gametogenesis in Clytia hemisphaerica jellyfish and examine the role of Spo11 using CRISPR-Cas9 mutants, generated clonally from F0 polyp colonies. Spo11 mutant females fail to assemble synaptonemal complexes and chiasmata, such that homologous chromosome pairs disperse during oocyte growth. Subsequent meiotic divisions are abnormal but produce viable progeny. Clytia thus shares an ancient eukaryotic dependence of synapsis and chromosome segregation on Spo11-generated DSBs. It provides a valuable additional experimental model for dissecting meiotic mechanisms during animal gametogenesis, and for building a comparative framework for distinguishing evolutionarily conserved versus flexible features of meiosis.

Lipases and carboxylesterases are involved in interspecific pheromone differences between two moth species

Moth sex pheromones are a classical model for studying sexual selection. Females produce a species-specific pheromone blend that attracts males. Revealing the enzymes involved in the interspecific variation in blend composition is key for understanding the evolution of these sexual communication systems. The nature of the enzymes involved in the variation of acetate esters, which are prominent compounds in moth pheromone blends, remains unclear. We identified enzymes involved in acetate metabolism in two closely related species: Heliothis (Chloridea) subflexa and H. (C.) virescens, which differ in production of acetate esters. Through comparative transcriptomic analyses and CRISPR/Cas9 knockouts, we showed that two lipases and two esterases induce lower levels of acetate esters in female pheromones. To place our findings in an evolutionary context, we explored the molecular evolution of related lipases and esterases in Lepidoptera. Together, our results show that lipases and carboxylesterases are unexpectedly involved in tuning Lepidoptera pheromones composition.

Comparing Deep Neural Networks and Gradient Boosting for Pneumonia Detection Using Chest X-Rays

In recent years, with the development of computational power and the explosion of data available for analysis, deep neural networks, particularly convolutional neural networks, have emerged as one of the default models for image classification, outperforming most of the classical machine learning models in this task. On the other hand, gradient boosting, a classical model, has been widely used for tabular structure data and leading data competitions, such as those from Kaggle. In this study, the authors compare the performance of deep neural networks with gradient boosting models for detecting pneumonia using chest x-rays. The authors implement several popular architectures of deep neural networks, such as Resnet50, InceptionV3, Xception, and MobileNetV3, and variants of a gradient boosting model. The authors then evaluate these two classes of models in terms of prediction accuracy. The computation in this study is done using cloud computing services offered by Google Colab Pro.

Study of hepatotoxins influence in vitro on basic biochemical indicators of liver functional state

An antimicrobial drug of the fluoroquinolone group, ciprofloxacin, is widely used in Ukraine. However, some researchers have noted the probable hepatotoxicity of this drug with prolonged use or use of high doses of ciprofloxacin. The aim of this study was to compare the effects of carbon tetrachloride, as a classical model of hepatocyte injury, with the effects of ciprofloxacin. The aim of the study was to investigate the biochemical parameters of the liver when simulating toxic damage to hepatocytes with carbon tetrachloride or ciprofloxacin. Materials and methods. The study was carried out on isolated rat hepatocytes, in whose culture medium carbon tetrachloride or ciprofloxacin was added. After incubation in the supernatant and cell homogenate, the activities of marker enzymes of cytolysis were determined: ALT, AST, γ-GTP, LF, LDH, DC and MDA. Results. The introduction of ciprofloxacin into the culture of hepatocytes at a concentration of LC50 caused changes in biochemical parameters similar to those caused by carbon tetrachloride. Thus, an increase in ALT, AST, γ-GTP, LF, LDH, DC and MDA was observed when CCl4 or ciprofloxacin was added to the culture. Conclusion. Incubation of rat hepatocytes with carbon tetrachloride or ciprofloxacin caused an increase in the level of enzymes and lipid peroxidation products. These parameters are indicators of gross changes in cells, which are the result of impaired keto acid formation, impaired redox reactions, impaired glycogen production

Genomic environments scale the activities of diverse core promoters

A classical model of gene regulation is that enhancers provide specificity whereas core promoters provide a modular site for the assembly of the basal transcriptional machinery. However, examples of core promoter specificity have led to an alternate hypothesis in which specificity is achieved by core promoters with different sequence motifs that respond differently to genomic environments containing different enhancers and chromatin landscapes. To distinguish between these models, we measured the activities of hundreds of diverse core promoters in four different genomic locations and, in a complementary experiment, six different core promoters at thousands of locations across the genome. Although genomic locations had large effects on expression, the intrinsic activities of different classes of promoters were preserved across genomic locations, suggesting that core promoters are modular regulatory elements whose activities are independently scaled up or down by different genomic locations. This scaling of promoter activities is nonlinear and depends on the genomic location and the strength of the core promoter. Our results support the classical model of regulation in which diverse core promoter motifs set the intrinsic strengths of core promoters, which are then amplified or dampened by the activities of their genomic environments.

MORPHOLOGICAL MONITORING OF EXPERIMENTAL LIVER FIBROSIS IN RATS

Background. Though thioacetamide (TAA)-induced liver fibrosis (LF) is recognized as a classical model of toxic liver damage, there is no literature data on the description of its successive stages of histological and ultrastructural changes in various cell populations involved in fibrosis. Objective. To conduct morphological monitoring of fibrosis formation in the liver of rats using the TAA model of LF based on histological and ultrastructural changes in hepatocytes and perisinusoidal lipocytes (HSC). Material and methods. The experiment was carried out on 18 sexually mature male rats. LF was modeled by intraperitoneal injection of 2% TAA solution at a dose of 10 ml / kg every other day. Light microscopy of semi-thin sections of the liver was performed, as well as electron microscopy of ultrathin sections. Results. The study of semi-thin sections of rat liver tissue from the control group showed a normal architecture of the parenchyma, a large number of HSCs containing large lipid droplets ("resting" phenotype), a very small amount of cytoplasmic matrix poor in membrane organelles. In the animals that were receiving TAA for 4 weeks, a mesenchymalepithelial transition of HSCs from the "resting" type to a fibrogenic state (fibrogenic phenotype) was recorded, that was accompanied by a gradual decrease in the number of retinol-containing drops and the appearance of fibroblastlike cells (FLC) in HSCs. In the animals, that were receiving TAA for 12 weeks, the pool of fibrogenic cells in the liver increased, a mesothelial-mesenchymal transition occurred, characterized by the mesothelial cell migration deeper into the parenchyma and their acquisition of a mesenchymal phenotype. Lipid containing activated FLC were also found in fibrous tissue around the central vein. Foci of hepatic tissue destruction caused by necrosis and apoptosis of hepatocytes were much more common. Conclusions. Administration of TAA induces liver fibrosis while histological and ultrastructural monitoring of the state of hepatocytes and HSCs allows to monitor all stages of fibrosis, clarifying the mechanisms of damage to intracellular organelles and variants of hepatocyte death. This model of LF in rats can be used to test new antifibrotic drugs.

Electrodeposition of Cu-Ag Alloy Films at n-Si(001) and Polycrystalline Ru Substrates

Electrodeposition of Cu-Ag films from acidic sulfate bath was conducted at n-Si(001) and polycrystalline Ru substrates. Significant nucleation overpotential of 0.4 V is observed with the Cu-Ag bath at n-Si(001) substrate, whereas the electrodeposition of Cu-Ag at Ru substrate is influenced by Ru oxides at the surface. Incomplete coverage of Si substrate by Cu-Ag deposit was observed from the deposition systems without Ag(I), or with 0.1 mM Ag(I), comparing with the compact Cu-Ag film obtained with the deposition bath containing 0.01 mM Ag(I). Layered and faceted Cu-Ag deposit was observed at small Cu deposition overpotential with the Ru substrate. Phase composition of the Cu-Ag deposits at n-Si(001) substrate from electrolyte with various Ag(I) concentrations is examined by XRD. Limited solubility of Ag (0.4 at.%) was observed in fcc-Cu until phase separation occurs. The classical model for nucleation kinetics in electrodeposition was used to examine the potentiostatic transients of the Cu-Ag electrodeposition at n-Si(001) substrate.

Energy exchange calculations in a simple mechanical system to investigate the origin of friction

The microscopic origin of friction is an important topic in science and technology. To date, noteworthy aspects of it remain unsolved. In an effort to shed some light on the possible mechanisms that could give rise to the macroscopic emergence of friction, a very simple 1D system of two particles is considered, one of them is free but moving with an initial velocity, and the other confined by a harmonic potential. The two particles interact via a repulsive Gaussian potential. While it represents in a straightforward manner a tip substrate system in the real world, no analytic solutions can be found for its motion. Because of the interaction, the free particle (tip) may overcome the bound particle (substrate) losing part of its kinetic energy. We solve Newton’s equations of the two particles numerically and calculate the net exchange of energy in the asymptotic state in terms of the relevant parameters of the problem. The effective dissipation that emerges from this simple, classical model with no ad hoc terms shows, surprisingly, a range of rich, nontrivial, behavior. We give theoretical reasoning which provides a satisfactory qualitative description. The essential ingredient of that reasoning is that the transfer of energy from the incoming particle to the confined one can be regarded as the source of the emergent dissipation force the friction experienced by the incoming particle.