Deepfake technology, driven by Generative Adversarial Networks (GANs), poses significant risks to privacy and societal security. Existing detection methods are predominantly passive, focusing on post-event analysis without preventing attacks. To address this, we propose an active defense method based on low-frequency perceptual perturbations to disrupt face-swapping manipulation, reducing the performance and naturalness of generated content. Unlike prior approaches that used low-frequency perturbations to impact classification accuracy, our method directly targets the generative process of deepfake techniques.We combine frequency and spatial domain features to strengthen defenses. By introducing artifacts through low-frequency perturbations while preserving high-frequency details, we ensure the output remains visually plausible. Additionally, we design a complete architecture featuring an encoder, a perturbation generator, and a decoder, leveraging discrete wavelet transform (DWT) to extract low-frequency components and generate perturbations that disrupt facial manipulation models. Experiments on CelebA-HQ and LFW demonstrate significant reductions in face-swapping effectiveness, improved defense success rates, and preservation of visual quality.

Abstract Background RNA methylation (RM) regulates gene expression regulation, RNA stability, and protein translation. Accurate prediction of RM modification sites is essential for understanding their biological functions. However, existing wet-lab detection techniques face challenges including operational complexity and high costs. Deep learning (DL) methods have been applied to this task. However, existing methods show performance degradation with smaller training datasets. For instance, the Bidirectional Gated Recurrent Unit (BGRU) demonstrates substantial performance degradation. Contrastive Learning Network (CNN) can extract local pattern features but learns overly specific patterns with sample-limited data, resulting in poor feature generalization. Bidirectional Long Short-Term Memory (BiLSTM) excels at modeling long-range dependencies but cannot sufficiently learn gating mechanism parameters to capture effective sequence representations with limited samples. Transformer processes sequences in parallel and captures global dependencies through self-attention, but its quadratic computational complexity and large parameter count make it prone to overfitting on small datasets. Current DL methods show reduced performance when training data is limited. Results This study proposes a Multi-view Contrastive Learning with CNN-BiLSTM-Attention (MCLCBA) framework for RM modification site prediction. The multi-view approach comprises a primary view and auxiliary view, where the primary view utilizes DNA Bidirectional Encoder Representations from Transformers (DNABERT) to extract sequence contextual features, and the auxiliary view employs Chaos Game Representation (CGR) to extract structural features. Feature extraction includes four components: data augmentation, multi-view encoders, projection heads, and contrastive loss functions. By implementing dual differential data augmentation strategies and constructing multi-view network architectures for feature processing and fusion, the model learns discriminative feature representations invariant to data augmentation through maximizing positive sample similarity while minimizing negative sample similarity. This effectively addresses sample-limited feature learning scenarios. Experimental results on the sample-limited m 7 G dataset demonstrate that MCLCBA achieves AUROC and AUPRC of 85.64% and 86.94%, respectively, improving upon existing methods by 5–6% in both metrics. Conclusions Through multi-view contrastive learning, MCLCBA provides an approach for RM sites under sample-limited scenarios.

The rise in wearable devices has led to the generation of a large amount of unlabeled electrocardiogram (ECG) data. Effectively utilising this data has been a challenge. One approach to address this issue is contrastive learning. However, most existing contrastive learning methods based on data augmentation primarily utilize the augmented electro-cardiogram (ECG) signals for comparison. These approaches have certain drawbacks. While the augmented ECG signals may help highlight certain features, they could also potentially mask important variations in the original signal, leading to the loss of crucial information. Moreover, solely relying on augmented data may lead to the model relying too heavily on specific variation patterns, overlooking the authentic features present in the original signal, resulting in poor generalization and robustness for downstream tasks. To address this issue, we propose a multi-lead electrocardiogram signal classification method based on temporal and multi-view contrastive learning. The method utilizes the time invariance of ECG signals and the multi-view information provided by the original ECG signals and augmented signals for contrastive learning. Through this approach, it not only addresses the limitations of relying solely on augmented signals but also leverages the prior knowledge that ECG signal categories remain stable over short periods of time to fully model the time context of ECG signals, capturing dynamic features. The integration of the temporal context comparison module and the multi-view comparison module significantly enhances the performance of downstream classification tasks. The outcomes of our experiments show that our method performs better on four datasets than previous approaches, which even exceeds supervised performance. When we pretrain on the SPH dataset and fine-tune on the PTB-XL dataset, our approach shows the best performance. Specifically, our method's AUROC exceeds that of the best baseline model by 4.9% and surpasses supervised performance by 1.3%.

Molecular representation learning is a fundamental problem in the field of drug discovery and molecular science. Whereas incorporating molecular 3D information in the representations of molecule seems beneficial, which is related to computational chemistry with the basic task of predicting stable 3D structures (conformations) of molecules. Existing machine learning methods either rely on 1D and 2D molecular properties or simulate molecular force field to use additional 3D structure information via Hamiltonian network. The former has the disadvantage of ignoring important 3D structure features, while the latter has the disadvantage that existing Hamiltonian neural network must satisfy the “canonial” constraint, which is difficult to be obeyed in many cases. In this paper, we propose a novel plug-and-play architecture LagNet by simulating molecular force field only with parameterized position coordinates, which implements Lagrangian mechanics to learn molecular representation by preserving 3D conformation without obeying any additional restrictions. LagNet is designed to generate known conformations and generalize for unknown ones from molecular SMILES. Implicit positions in LagNet are learned iteratively using discrete-time Lagrangian equations. Experimental results show that LagNet can well learn 3D molecular structure features, and outperforms previous state-of-the-art baselines related molecular representation by a significant margin.

OBJECTIVE: Due to the lack of fine-grained labels, current research can only evaluate the signal quality at a coarse scale. This article proposes a weakly supervised fine-grained electrocardiogram (ECG) signal quality assessment method, which can produce continuous segment-level quality scores with only coarse labels. METHODS: A novel network architecture, i.e. FGSQA-Net, is developed for signal quality assessment, which consists of a feature shrinking module and a feature aggregation module. Multiple feature shrinking blocks, which combine residual CNN block and max pooling layer, are stacked to produce a feature map corresponding to continuous segments along the spatial dimension. Segment-level quality scores are obtained by feature aggregation along the channel dimension. RESULTS: The proposed method was evaluated on two real-world ECG databases and one synthetic dataset. Our method produced an average AUC value of 0.975, which outperforms the state-of-the-art beat-by-beat quality assessment method. The results are visualized for 12-lead and single-lead signals over a granularity from 0.64 to 1.7 seconds, demonstrating that high-quality and low-quality segments can be effectively distinguished at a fine scale. CONCLUSION: FGSQA-Net is flexible and effective for fine-grained quality assessment for various ECG recordings and is suitable for ECG monitoring using wearable devices. SIGNIFICANCE: This is the first study on fine-grained ECG quality assessment using weak labels and can be generalized to similar tasks for other physiological signals.

N6-methyladenosine (m6A) has been shown to play crucial roles in RNA metabolism, physiology, and pathological processes. However, the specific regulatory mechanisms of most methylation sites remain uncharted due to the complexity of life processes. Biological experimental methods are costly to solve this problem, and computational methods are relatively lacking. The discovery of local co-methylation patterns (LCPs) of m6A epi-transcriptome data can benefit to solve the above problems. Based on this, we propose a novel biclustering algorithm based on the beta distribution (BDBB), which realizes the mining of LCPs of m6A epi-transcriptome data. BDBB employs the Gibbs sampling method to complete parameter estimation. In the process of modeling, LCPs are recognized as sharp beta distributions compared to the background distribution. Simulation study showed BDBB can extract all the three actual LCPs implanted in the background data and the overlap conditions between them with considerable accuracy (almost close to 100%). On MeRIP-Seq data of 69,446 methylation sites under 32 experimental conditions from 10 human cell lines, BDBB unveiled two LCPs, and Gene Ontology (GO) enrichment analysis showed that they were enriched in histone modification and embryo development, etc. important biological processes respectively. The GOE_Score scoring indicated that the biclustering results of BDBB in the m6A epi-transcriptome data are more biologically meaningful than the results of other biclustering algorithms.

Multivariate time series (MTS) prediction aims at predicting future time series by extracting multiple forms of dependencies of past time series. Traditional prediction methods and deep learning-based prediction methods focus on extracting the dynamic relationships of certain aspects of MTS, especially the temporal characteristics, often neglecting the spatial and temporal dynamic correlations of MTS. Inspired by convolution neural network (CNN) and attention mechanism, this paper proposes a convolution LSTM network model based on MTS prediction with two-stage attention. Specifically, we first propose a new MTS preprocessing method to perform convolution operations better. Then convolution layer extracts spatial correlation of MTS and LSTM model extracts temporal correlation. It is worth mentioning that the combination of attention mechanism and LSTM can effectively solve the problem of insufficient time dependency in MTS prediction. In addition, dual-stage attention mechanism can effectively eliminate irrelevant information, select the relevant exogenous sequence, give it higher weight, and increase the past value of the target sequence to further eliminate irrelevant information. Finally, the MTS spatio-temporal correlation is extracted to improve the prediction accuracy, and the model is interpreted. Experimental results show that the model has broad application prospects. Experiments based on typical datasets of finance, environment, and energy determine the optimal window size and hidden size of the prediction, and demonstrate that the model achieves the state-of-the-art effect compared to the other four deep learning models. On top of that, the model is not only suitable for single-step prediction of MTS, but also suitable for multistep prediction of time step in a certain range.