The rational utilization of multimodal spatial transcriptomics (ST) data enables accurate identification of spatial domains, which is essential for investigating cellular structure and functions. In this study, we proposed SpaConTDS, a novel framework that integrates reinforcement learning with self-supervised multimodal contrastive learning. SpaConTDS generates positive and negative samples through data augmentation and a pseudo-label tuple perturbation strategy, enabling the learning of fused representations that capture global semantics and cross-modal interactions. The model's hyper-parameters are dynamically optimized using reinforcement learning. Extensive experiments across various resolutions and platforms demonstrate that SpaConTDS achieves state-of-the-art accuracy in spatial domain identification and outperforms existing methods in downstream tasks such as denoising, trajectory inference, and UMAP visualization. Moreover, SpaConTDS effectively integrates multiple tissue sections and corrects batch effects without requiring prior alignment. Compared to existing approaches, SpaConTDS offers more robust fused representations of multimodal data, providing researchers with a flexible and powerful tool for a wide range of spatial transcriptomics analyses.

Hybrid tabular-textual question answering (QA) typically requires numerical reasoning over heterogeneous data, with the reasoning program first generated and then executed to obtain the final answer. However, in most existing hybrid QA benchmarks, each step only relies on the numbers in the input or the calculation result of the preceding step. Questions that require long-distance numerical reasoning (where the calculation of a step relies on the results calculated a number of steps previously) are rare. However, they may be required to solve some complicated financial problems; these questions require a more complex model capable of capturing long-distance dependencies. To study more challenging hybrid tabular-textual QA, we construct a new large-scale hybrid tabular-textual dataset, COLD-QA, COmplex Long-Distance numerical reasoning Question Answering dataset. We also conduct extensive experiments with multiple baselines. The COLD-QA dataset is significantly more difficult than previous work, according to experiment results.

MOTIVATION: Single cell RNA sequencing (scRNA-seq) technique enables the transcriptome profiling of hundreds to ten thousands of cells at the unprecedented individual level and provides new insights to study cell heterogeneity. However, its advantages are hampered by dropout events. To address this problem, we propose a Blockwise Accelerated Non-negative Matrix Factorization framework with Structural network constraints (BANMF-S) to impute those technical zeros. RESULTS: BANMF-S constructs a gene-gene similarity network to integrate prior information from the external PPI network by the Triadic Closure Principle and a cell-cell similarity network to capture the neighborhood structure and temporal information through a Minimum-Spanning Tree. By collaboratively employing these two networks as regularizations, BANMF-S encourages the coherence of similar gene and cell pairs in the latent space, enhancing the potential to recover the underlying features. Besides, BANMF-S adopts a blocklization strategy to solve the traditional NMF problem through distributed Stochastic Gradient Descent method in a parallel way to accelerate the optimization. Numerical experiments on simulations and real datasets verify that BANMF-S can improve the accuracy of downstream clustering and pseudo-trajectory inference, and its performance is superior to seven state-of-the-art algorithms. AVAILABILITY: All data used in this work are downloaded from publicly available data sources, and their corresponding accession numbers or source URLs are provided in Supplementary File Section 5.1 Dataset Information. The source codes are publicly available in Github repository https://github.com/jiayingzhao/BANMF-S.

Continual learning is gaining special attention in remaining useful life (RUL) prediction of machinery recently, which enables deep prognostics networks to use incremental samples to progressively improve network performance without laborious retraining. Nonetheless, current studies exhibit several constraints: 1) An explicit mechanism is lacking in preventing the loss of pivotal memories after multiple continual learning stages. 2) A sampling-enhanced replay technique is lacking for continual learning-based RUL prediction. To address the abovementioned limitations, this article proposes a complementary continual learning framework for RUL prediction of machinery, which contains two novel characteristics, i.e., long-term potentiation and associative replay. These two characteristics are complementary and coenhanced. The long-term potentiation focuses on multistage continual learning, which is able to prevent deep prognostics networks from forgetting the formerly learned degradation knowledge. The associative replay pays attention to each new continual learning stage, which is able to consolidate typical degradation knowledge into new network learning. The proposed framework is verified using run-to-failure datasets from rolling element bearings, and the framework is also compared with some state-of-the-art methods. Experimental results indicate that the proposed framework can possess lower forgetting and achieve better prognostics performance reinforcement during continual learning.

Spatial transcriptomics is a rapidly growing field that aims to comprehensively characterize tissue organization and architecture at single-cell or sub-cellular resolution using spatial information. Such techniques provide a solid foundation for the mechanistic understanding of many biological processes in both health and disease that cannot be obtained using traditional technologies. Several methods have been proposed to decipher the spatial context of spots in tissue using spatial information. However, when spatial information and gene expression profiles are integrated, most methods only consider the local similarity of spatial information. As they do not consider the global semantic structure, spatial domain identification methods encounter poor or over-smoothed clusters. We developed ConSpaS, a novel node representation learning framework that precisely deciphers spatial domains by integrating local and global similarities based on graph autoencoder (GAE) and contrastive learning (CL). The GAE effectively integrates spatial information using local similarity and gene expression profiles, thereby ensuring that cluster assignment is spatially continuous. To improve the characterization of the global similarity of gene expression data, we adopt CL to consider the global semantic information. We propose an augmentation-free mechanism to construct global positive samples and use a semi-easy sampling strategy to define negative samples. We validated ConSpaS on multiple tissue types and technology platforms by comparing it with existing typical methods. The experimental results confirmed that ConSpaS effectively improved the identification accuracy of spatial domains with biologically meaningful spatial patterns, and denoised gene expression data while maintaining the spatial expression pattern. Furthermore, our proposed method better depicted the spatial trajectory by integrating local and global similarities.

Inference of gene regulatory network (GRN) from gene expression profiles has been a central problem in systems biology and bioinformatics in the past decades. The tremendous emergency of single-cell RNA sequencing (scRNA-seq) data brings new opportunities and challenges for GRN inference: the extensive dropouts and complicated noise structure may also degrade the performance of contemporary gene regulatory models. Thus, there is an urgent need to develop more accurate methods for gene regulatory network inference in single-cell data while considering the noise structure at the same time. In this paper, we extend the traditional structural equation modeling (SEM) framework by considering a flexible noise modeling strategy, namely we use the Gaussian mixtures to approximate the complex stochastic nature of a biological system, since the Gaussian mixture framework can be arguably served as a universal approximation for any continuous distributions. The proposed non-Gaussian SEM framework is called NG-SEM, which can be optimized by iteratively performing Expectation-Maximization algorithm and weighted least-squares method. Moreover, the Akaike Information Criteria is adopted to select the number of components of the Gaussian mixture. To probe the accuracy and stability of our proposed method, we design a comprehensive variate of control experiments to systematically investigate the performance of NG-SEM under various conditions, including simulations and real biological data sets. Results on synthetic data demonstrate that this strategy can improve the performance of traditional Gaussian SEM model and results on real biological data sets verify that NG-SEM outperforms other five state-of-the-art methods.

Advances in single-cell RNA sequencing (scRNA-seq) technology provide an unbiased and high-throughput analysis of each cell at single-cell resolution, and further facilitate the development of cellular heterogeneity analysis. Despite the promise of scRNA-seq, the data generated by this method are sparse and noisy because of the presence of dropout events, which can greatly impact downstream analyses such as differential gene expression, cell type annotation, and linage trajectory reconstruction. The development of effective and robust computational methods to address both dropout and clustering are thus urgently needed. In this study, we propose a flexible, accurate two-stage algorithm for single cell heterogeneity analysis via hierarchical clustering based on an optimal imputation strategy, called scHOIS. At the first stage, masked non-negative matrix factorization is applied to approximate the original observed scRNA-seq data, with optimal rank determined by variance analysis. At the second stage, hierarchical clustering is applied to group the imputed cells using Pearson correlation to measure similarity, with the optimal number of clusters determined by integrating three classical indexes. We performed extensive experiments on real-world datasets, which showed that scHOIS effectively and robustly distinguished cellular differences and that the clustering performance of this algorithm was superior to that of other state-of-the-art methods.

The identification of drug side-effects is considered to be an important step in drug design, which could not only shorten the time but also reduce the cost of drug development. In this paper, we investigate the relationship between the potential side-effects of drug candidates and their chemical structures. The preliminary Regularized Regression (RR) model for drug side-effects prediction has promising features in the efficiency of model training and the existence of a closed form solution. It performs better than other state-of-the-art methods, in terms of minimum accuracy and average accuracy. In order to dig inside how drug structure will associate with side effect, we further propose weighted GTS (Generalized T-Student Kernel: WGTS) SVM model from a structural risk minimization perspective. The SVM model proposed in this paper provides a better understanding of drug side-effects in the process of drug development. The usefulness of the WGTS model lies in the superior performance in a cross validation setting on 888 approved drugs with 1385 side-effects profiling from SIDER database. This work is expected to shed light on intriguing studies that predict potential un-identifying side-effects and suggest how we can avoid drug side-effects by the removal of some distinguished chemical structures.

This paper studies the problem of exactly identifying the structure of a probabilistic Boolean network (PBN) from a given set of samples, where PBNs are probabilistic extensions of Boolean networks. Cheng et al. studied the problem while focusing on PBNs consisting of pairs of AND/OR functions. This paper considers PBNs consisting of Boolean threshold functions while focusing on those threshold functions that have unit coefficients. The treatment of Boolean threshold functions, and triplets and -tuplets of such functions, necessitates a deepening of the theoretical analyses. It is shown that wide classes of PBNs with such threshold functions can be exactly identified from samples under reasonable constraints, which include: 1) PBNs in which any number of threshold functions can be assigned provided that all have the same number of input variables and 2) PBNs consisting of pairs of threshold functions with different numbers of input variables. It is also shown that the problem of deciding the equivalence of two Boolean threshold functions is solvable in pseudopolynomial time but remains co-NP complete.

Snow detection and removal from video images is very challenging. Normally the snowflakes affect only on a very small region of an image, hence the confusion to determine which region should be considered and which one should not. In this paper, a frame difference method with five successive frames is first presented to detect the snow pixels from image background, but the method didn't work well in the case of heavy snow. Then a new technique has been implemented which uses the L0gradient minimization approach to remove the snow pixels. This technique can control how many non-zero gradients are resulted in the image, and is independent of local features, but instead locates important edges globally. These salient edges are preserved and the low amplitude and insignificant details are diminished. The snow pixels are then removed in this way. Experimental results show that this method is a highly efficient algorithm even under heavy snow conditions, while preserving the details of the image.

Positive semi-definiteness is a critical property in Support Vector Machine (SVM) methods to ensure efficient solutions through convex quadratic programming. In this paper, we introduce a projection matrix on indefinite kernels to formulate a positive semi-definite one. The proposed model can be regarded as a generalized version of the spectrum method (denoising method and flipping method) by varying parameter λ. In particular, our suggested optimal λ under the Bregman matrix divergence theory can be obtained using unconstrained optimization. Experimental results on 4 real world data sets ranging from glycan classification to cancer prediction show that the proposed model can achieve better or competitive performance when compared to the related indefinite kernel methods. This may suggest a new way in motif extractions or cancer predictions.

We study the number of samples required to uniquely determine the structure of a probabilistic Boolean network (PBN), where PBNs are probabilistic extensions of Boolean networks. We show via theoretical analysis and computational analysis that the structure of a PBN can be exactly identified with high probability from a relatively small number of samples for interesting classes of PBNs of bounded indegree. On the other hand, we also show that there exist classes of PBNs for which it is impossible to uniquely determine the structure of a PBN from samples.

Boolean network (BN) is a popular mathematical model for revealing the behavior of a genetic regulatory network, and observability plays a vital role in understanding the underlying network feature. However, the observability of attractor cycles, which is an interesting and important problem, has not been addressed in the literature. In this paper, we first proposed a novel problem on attractor observability in BNs. Identification of the minimum set of consecutive nodes can be used to determine uniquely the attractor cycle from the others in the network. We then develop a linear-time algorithm to identify the desired set of nodes. The proposed approaches are demonstrated and verified by numerical examples. The computational results are given to illustrate both the efficiency and effectiveness of our proposed methods.

Facing the important issues of safety analysis and assessment for the train service state, an online quantified safety assessment method based on the safety region estimation and hybrid intelligence technologies was proposed in this paper. First, the previous researches on the safety analysis and assessment were briefly reviewed for the train itself and its key equipment, and the existential problems were further pointed out. Then, using the safety monitoring data and the safety region estimation theory, a new online safety assessment method with data-driven was put forward, which was followed by a detailed description of the concrete implementation steps including the EMD (Local Mean Decomposition) and EM (Energy Moment) based safety risk evaluation index selection, Interval Type 2 Fuzzy C-Means (IT2FCM) clustering based safety region boundary calculation modeling and safety risk grading. Finally, in order to verify its performance through experiments, the above method was applied in analyzing and evaluating service states of the rolling bearings, the key equipment of the train, on the basis of mass field data. The experimental results indicate that this method is valid.

This paper presents a new approach for on-line collision-free path planning of a two-arm manipulator system, integrated in an on-line task-level planning system. For service and assembly tasks, pick and place operations are requested asynchronously by the action sequence planning. Collision-free paths for independent tasks are planning using a 2D geometric model in consideration of the swept regions by the robot arms and the pay loads during their motions. A dynamic, deadlock-free scheduling concept coordinates the robot motions in the case that a collision-free path for one arm can not be found at moment. The on-line path planning for two-arm cooperations for exchanging and regrasping parts incorporates an off-line connectivity analysis, avoiding both collisions and kinematic restrictions. The implemented on-line planning system can generate collision-free path for one manipulator while the other is moving. The time needed for motion planning is in the average case shorter and in the worst case comparable with that needed for motion execution. Experiments have been successfully conducted with the mobile two-arm robot KAMRO at the author's institute.

A fast method for computing the collision vector for online collision detection with a multi-arm robot is presented. Manipulators and obstacles are modeled by sets of convex polytopes. Known distance algorithms serve as a foundation. To speed up the collision detection dynamic obstacles are approximated by geometric primitives and organized in hierarchies. Online, the dynamic hierarchies are adjusted to the current arm configuration. A comparison with previous methods showed an increased acceleration of the computations.>