Histone modifications play a crucial role in transcriptional regulation and are essential targets for genome annotation and gene expression modeling. However, experimentally profiling histone modifications across species, tissues, and cellular states is costly and often impractical. While numerous deep learning models have been proposed to predict histone modifications, most operate at relatively low resolution (128 bp), limiting their utility in fine-scale genomic analysis. In this study, we present MambaHM (Mamba for predicting Histone Modifications), a novel deep learning framework built upon Mamba architecture. MambaHM integrates DNA sequence features with chromatin accessibility data (ATAC-seq) to predict ten histone modifications. Leveraging the linear computational complexity of Mamba's state-space modeling, our model avoids excessive compression of feature length during embedding. This enables the model to retain sufficient contextual information while simultaneously achieving 16-bp resolution, thereby enhancing the granularity and accuracy of prediction. Experiments demonstrate that MambaHM achieves a mean Pearson correlation of 0.836 (± 0.063) on the K562 cell line test set. Furthermore, the model generalizes well across cell types, tissues, and species, with a mean cross-context correlation of$0.557 (\pm 0.105)$, approaching the reliability of experimental assays. Compared to state-of-the-art models, MambaHM achieves a$\mathbf{7. 0 3 3 \%}(\mathbf{\pm 0. 0 4 1})$improvement in cross-cell-type performance and over a 58.978 % (877 bp) improvement in prediction deviation evaluation for peak calling. Overall, MambaHM provides a powerful, cost-effective, and fineresolution tool for histone modification prediction, offering precise epigenomic references for downstream analysis and potential applications in drug discovery. The source code is available at https://github.com/zhichunlizzx/MambaHM.

This letter studies the issue of robust multitask distributed estimation under the error-in-variable (EIV) model where input noise and output impulsive noise are considered. In such cases, existing distributed algorithms suffer from severe performance degradation. To tackle this problem, a robust multitask diffusion bias-compensated least mean M-estimate (R-MD-BCLMM) is proposed. We adopt a new real-time input noise variance estimation method which utilizes piecewise linearity of the modified Huber function to resist input noises. To further improve network information exchange capability and estimation performance, a robust spatial average combination based multitask adaptive clustering strategy is proposed. Finally, simulations demonstrate that the proposed R-MD-BCLMM algorithm outperforms some state-of-the-art distributed algorithms.

Histone modifications (HMs) play a critical role in various biological processes, but annotating histone modifications across different cell types using experimental methods alone is extremely challenging. Although many deep learning methods have been developed to predict histone modifications, most rely solely on DNA sequences and do not incorporate novel cell-specific features. In this study, we propose KAS-former, a transformer-based model that integrates DNA sequences with cell-specific features derived from KAS-seq data, enabling effective prediction of histone modifications. Leveraging this transformer architecture coupled with dilated convolution, KAS-former achieves a broad receptive field, effectively capturing cell type-specific specificity from KAS-seq data. Our results demonstrate that KAS-former achieves high accuracy in predicting histone modifications across multiple cell types and shows strong potential for transcription factor prediction. By capturing cell-specific features, this approach not only improves the accuracy of histone modification predictions but also offers valuable insights into the interplay between histone modifications and transcription regulation. The code for KAS-former is available on GitHub at https://github.com/wzhy2000/KAS-former.

Histone modifications (HMs) are pivotal in various biological processes, including transcription, replication, and DNA repair, significantly impacting chromatin structure. These modifications underpin the molecular mechanisms of cell-type-specific gene expression and complex diseases. However, annotating HMs across different cell types solely using experimental approaches is impractical due to cost and time constraints. Herein, we present dHICA (deep histone imputation using chromatin accessibility), a novel deep learning framework that integrates DNA sequences and chromatin accessibility data to predict multiple HM tracks. Employing the transformer architecture alongside dilated convolutions, dHICA boasts an extensive receptive field and captures more cell-type-specific information. dHICA outperforms state-of-the-art baselines and achieves superior performance in cell-type-specific loci and gene elements, aligning with biological expectations. Furthermore, dHICA's imputations hold significant potential for downstream applications, including chromatin state segmentation and elucidating the functional implications of SNPs (Single Nucleotide Polymorphisms). In conclusion, dHICA serves as a valuable tool for advancing the understanding of chromatin dynamics, offering enhanced predictive capabilities and interpretability.

We study the problem of distributed estimation over wireless sensor networks (WSNs), where measurement noises and false data injection (FDI) attacks are considered. We propose the malicious node filtrating based secure diffusion bias compensated recursive least squares (MFS-dBCRLS) algorithm. To resist input noises, the BCRLS algorithm is introduced in local adaptation. To reduce the effect of network adversaries, a suitable time gating and a two-stage malicious node filtrating method are presented. Based on instantaneous state detection results in the first temporary filtrating stage, we permanently filtrate malicious neighbors in the second stage by proposing a new threshold test to decide final state. Besides, the selection range of the threshold is detailedly analyzed. Simulation results reveal communication-efficiency and robustness of our proposed method compared with some state-of-the-art algorithms under different FDI attacks.

In this letter, we study the problem of adaptive parameter estimation for distributed wireless sensor networks (WSNs), where the non-Gaussian impulsive noises in the communication links are considered. In such cases, if the traditional distributed collaborative strategy is still adopted, the estimation performance of the network will decline significantly. Aiming at this problem, we propose a new spatial-temporal minimum error random interaction strategy for the distributed estimation in the presence of impulsive link noises. Furthermore, the maximum correlation entropy criterion and the stochastic gradient descent method are used to update the combination factor so that the network has dynamic and real-time adaptability to the link noises. The proposed algorithm is also compared with the classic DLMS algorithm and a state-of-the-art algorithm designed for the impulsive link noises. Simulation results show that the proposed algorithm can not only reduce the network traffic effectively, but also be robust to the non-Gaussian link impulsive noises while maintaining its advantage over the non-cooperative algorithm and achieving the optimal estimation performance.

The problem of finite impulse response (FIR) filtering in errors‐in‐variables (EIV) system is studied. Due to the input noise, traditional recursive least‐squares (RLS) algorithms are biased in EIV system. Most existing bias‐compensated approaches are proposed in the case that both the input–output noises are white Gaussian random processes. However, taking account of the situation where the output is corrupted by coloured noise, there are rare existing algorithms work well. Two bias‐compensated RLS algorithms with acceptable computational complexity are proposed, which can obtain unbiased real‐time filtering in non‐stationary system when the input noise is white while the output noise is coloured. Under the assumption that the input signal is a coloured process, linear prediction technique is used to estimate the sample of the input signal. Exploiting the statistical properties of the cross‐correlation function between the least‐squares error and the forward/backward prediction error, the input noise variance can be estimated and the bias can be compensated. Simulation results illustrate the good performance of the proposed algorithms.