In insects and mammals, 3D genome topology has been linked to transcriptional states yet whether this link holds for other eukaryotes is unclear. Using both ligation proximity and fluorescence microscopy assays, we show that in Saccharomyces cerevisiae , Heat Shock Response ( HSR ) genes dispersed across multiple chromosomes and under the control of Heat Shock Factor (Hsf1) rapidly reposition in cells exposed to acute ethanol stress and engage in concerted, Hsf1-dependent intergenic interactions. Accompanying 3D genome reconfiguration is equally rapid formation of Hsf1-containing condensates. However, in contrast to the transience of Hsf1-driven intergenic interactions that peak within 10-20 min and dissipate within 1 h in the presence of 8.5% (v/v) ethanol, transcriptional condensates are stably maintained for hours. Moreover, under the same conditions, Pol II occupancy of HSR genes and RNA expression are detectable only later in the response and peak much later (>1 h). This contrasts with the coordinate response of HSR genes to thermal stress (39°C) where Pol II occupancy, transcription, intergenic interactions, and formation of Hsf1 condensates are all rapid yet transient (peak within 2.5-10 min and dissipate within 1 h). Therefore, Hsf1 forms condensates, restructures the genome and transcriptionally activates HSR genes in response to both forms of proteotoxic stress but does so with strikingly different kinetics. In cells subjected to ethanol stress, Hsf1 forms condensates and repositions target genes before transcriptionally activating them.

Bacterial biofilms are communities of bacteria usually attached to solid strata and often differentiated into complex structures. Communication across biofilms has been shown to involve chemical signaling and, more recently, electrical signaling in Gram positive biofilms. We report for the first time, community-level synchronized membrane potential dynamics in three-dimensional E. coli biofilms. Two hyperpolarization events are observed in response to light stress. The first requires mechanically sensitive ion channels (MscK, MscL and MscS) and the second needs the Kch-potassium channel. The channels mediated both local spiking of single E. coli biofilms and long-range coordinated electrical signaling in E. coli biofilms. The electrical phenomena are explained using Hodgkin-Huxley and 3D fire-diffuse-fire agent-based models. These data demonstrate that electrical wavefronts based on potassium ions are a mechanism by which signaling occurs in Gram negative biofilms and as such may represent a conserved mechanism for communication across biofilms.

Large code models (LCMs) have remarkably advanced the field of code intelligence. Despite their impressive capabilities, they still face practical employment challenges, such as high costs, limited accessibility of proprietary LCMs, and adaptability issues of ultra-large LCMs. These challenges highlight the critical need for more accessible, lightweight yet effective LCMs. In this paper, we propose IterKD, an Iter Knowledge Distillation framework, which aims at continually transferring the programming capabilities of larger, advanced LCMs (Teacher) to smaller, less powerful LCMs (Student). IterKD consists of three stages in one cycle: (1) Correct-and-Fault Knowledge Delivery stage aims at improving the student models capability to recognize errors while ensuring its basic programming skill during the knowledge transferring, which involves correctness-aware supervised learning and fault-aware contrastive learning methods. (2) Multi-view Feedback stage aims at measuring the quality of results generated by the student model from two views, including model-based and static tool-based measurement; (3) Feedback-based Knowledge Update stage aims at updating the student model adaptively by generating new questions at different difficulty levels, in which the difficulty levels are categorized based on the feedback in the last stage. By performing the training cycle iteratively, the student model is continuously refined through learning more advanced programming skills from the teacher model. Finally, based on the proposed IterKD framework, we develop a lightweight yet effective LCM, named IterCoder, which is built upon CodeLlama-7B. Experimental results show that IterCoder achieves a Pass@1 score of 65.2 on the HumanEval benchmark, outperforming over-30B-sized LCMs by an average of 47.51% and surpassing comparable-sized LCMs by an average of 118.47%.

In the field of Structural Dynamics, modal analysis is the foundation of System Identification and vibration-based inspection. However, despite their widespread use, current state-of-the-art methods for extracting modal parameters from multi-input multi-output (MIMO) frequency domain data are still affected by many technical limitations. Mainly, they can be computationally cumbersome and/or negatively affected by close-in-frequency modes. The Loewner Framework (LF) was recently proposed to alleviate these problems with the limitation of working with single-input data only. This work proposes a computationally improved version of the Lowner Framework, or iLF, to extract modal parameters more efficiently. Also, the proposed implementation is extended in order to handle multi-input multi-output data in the frequency domain. This new implementation is compared to state-of-the-art methods such as the frequency domain implementations of the Least Square Complex Exponential method and the Numerical Algorithm for Subspace State Space System Identification on numerical and experimental datasets. More specifically, a finite element model of a 3D Euler-Bernoulli beam is used for the baseline comparison and the noise robustness verification of the proposed MIMO iLF algorithm. Then, an experimental dataset from MIMO ground vibration tests of a trainer jet aircraft with over 91 accelerometer channels is chosen for the algorithm validation on a real-life application. Excellent results are achieved in terms of accuracy, robustness to noise, and computational performance by the proposed improved MIMO method, both on the numerical and the experimental datasets. The MIMO iLF MATLAB implementation is shared in the work supplementary material.

Semantic interaction (SI) in Dimension Reduction (DR) of images allows users to incorporate feedback through direct manipulation of the 2D positions of images. Through interaction, users specify a set of pairwise relationships that the DR should aim to capture. Existing methods for images incorporate feedback into the DR through feature weights on abstract embedding features. However, if the original embedding features do not suitably capture the users' task then the DR cannot either. We propose ImageSI, an SI method for image DR that incorporates user feedback directly into the image model to update the underlying embeddings, rather than weighting them. In doing so, ImageSI ensures that the embeddings suitably capture the features necessary for the task so that the DR can subsequently organize images using those features. We present two variations of ImageSI using different loss functions - ImageSI_MDS_Inverse, which prioritizes the explicit pairwise relationships from the interaction and ImageSI_Triplet, which prioritizes clustering, using the interaction to define groups of images. Finally, we present a usage scenario and a simulation based evaluation to demonstrate the utility of ImageSI and compare it to current methods.

In the Heliosphere, power-law particle distributions are observed e.g. upstream of interplanetary shocks, which can result from superdiffusive transport. This non-Gaussian transport regime may result from intermittent magnetic field structures. Recently, we showed that a L\'evy flight model reproduces the observed features at shocks: power-law distributions upstream and enhanced intensities at the shock. We extend the L\'evy flight model to study the impact of superdiffusive transport on particle acceleration at shocks. The acceleration time scale and spectral slope are compared to Gaussian diffusion and a L\'evy walk model. The fractional transport equation is solved by sampling the number density with the corresponding stochastic differential equation that is driven by an alpha-stable L\'evy distribution. For both Gaussian and superdiffusive transport we use a modified version of CRPropa 3.2. We obtain the number density and energy spectra for constant and energy-dependent anomalous diffusion and find, compared to the case of Gaussian diffusion, harder energy spectra at the shock as well as faster acceleration. The spectral slope is even harder than predicted for L\'evy walks. L\'evy flight models of superdiffusive transport lead to observed features in the Heliosphere. We further show that superdiffusive transport impacts the acceleration process by changing the probability to escape the shock. The flexibility of the L\'evy flight model allows for further studies in the future, taking the shock geometry and magnetic field structure into account.

Assay for Transposase-Accessible Chromatin sequencing (ATAC-Seq) is a widely used technique to explore gene regulatory mechanisms. For most ATAC-Seq data from healthy and diseased tissues such as tumors, chromatin accessibility measurement represents a mixed signal from multiple cell types. In this work, we derive reliable chromatin accessibility marker peaks and reference profiles for most non-malignant cell types frequently observed in the tumor micro-environment. We then integrate these data into the EPIC deconvolution framework (Racle et al ., 2017) to quantify cell-type heterogeneity in bulk ATAC-Seq data. Our EPIC-ATAC tool accurately predicts non-malignant and malignant cell fractions in tumor samples. When applied to a breast cancer cohort, EPIC-ATAC accurately infers the immune contexture of the main breast cancer subtypes.

We present a new approach for nonlinear dimensionality reduction, specifically designed for computationally expensive mathematical models. We leverage autoencoders to discover a one-dimensional neural active manifold (NeurAM) capturing the model output variability, plus a simultaneously learnt surrogate model with inputs on this manifold. The proposed dimensionality reduction framework can then be applied to perform outer loop many-query tasks, like sensitivity analysis and uncertainty propagation. In particular, we prove, both theoretically under idealized conditions, and numerically in challenging test cases, how NeurAM can be used to obtain multifidelity sampling estimators with reduced variance by sampling the models on the discovered low-dimensional and shared manifold among models. Several numerical examples illustrate the main features of the proposed dimensionality reduction strategy, and highlight its advantages with respect to existing approaches in the literature.

This is a continuation of our previous work (Advances in Mathematics 450 (2024), Paper No. 109768). In this paper, we characterize complete metrics with finite total Q-curvature as normal metrics for all dimensional cases. Secondly, we introduce another volume entropy to provide geometric information regarding complete non-normal metrics with finite total Q-curvature. In particular, we show that if the scalar curvature is bounded from below, the volume growth of such complete metrics is controlled.

This paper presents an approach to decomposing animated graphics into sprites, a set of basic elements or layers. Our approach builds on the optimization of sprite parameters to fit the raster video. For efficiency, we assume static textures for sprites to reduce the search space while preventing artifacts using a texture prior model. To further speed up the optimization, we introduce the initialization of the sprite parameters utilizing a pre-trained video object segmentation model and user input of single frame annotations. For our study, we construct the Crello Animation dataset from an online design service and define quantitative metrics to measure the quality of the extracted sprites. Experiments show that our method significantly outperforms baselines for similar decomposition tasks in terms of the quality/efficiency tradeoff.