Explore the vast range of titles available.

Accurate identification of the interactions between T‐cell receptors (TCRs) and antigenic epitopes presented by major histocompatibility complex (MHC) molecules is fundamental to advancing cancer immunotherapy. Nevertheless, predictive modeling of TCR–epitope binding remains challenging, as existing models struggle to generalize to unseen epitopes while often overlooking key physicochemical properties governing immune recognition. Here, a biologically informed vision‐guided deep learning framework (DAISY) is proposed for robust and interpretable TCR–epitope binding prediction. DAISY integrates hierarchical physicochemical features via a biologically inspired Condition‐Adaptive Fusion module, jointly modeling residue‐level spatial interactions and global biochemical context. DAISY consistently outperforms state‐of‐the‐art models across four generalization scenarios, notably improving ROC‐AUC by 11% and PR‐AUC by 16% over the strongest competitor in the most challenging Unseen‐Pair setting. DAISY also offers intuitive interpretability by localizing interaction‐relevant residues via Score‐CAM visualizations. Furthermore, its computational predictions are bridged to key immunological and clinical outcomes, demonstrating utility in correlating with T‐cell clonal expansion, identifying functional TCRs, and robustly forecasting patient survival. Together, DAISY can serve as a powerful tool for broad translational immunology and introduces a scalable modeling paradigm for next‐generation immune modeling. Integrating hierarchical physicochemical features via a bio‐inspired design, the DAISY framework robustly predicts T‐cell receptor binding, significantly outperforming state‐of‐the‐art models on unseen epitopes. It uniquely offers visual interpretability via Score‐CAM visualizations and profound clinical relevance, correlating with T‐cell expansion, enabling functional TCR screening, and defining functional biomarkers that forecast patient survival in immunotherapy.

Using citizen science resources and projects in university education is a burgeoning pedagogical tool that can promote real-world application of science, autonomous learning, and understanding of self-efficacy in science learning. In this case study, we examined several factors relating to self-efficacy and skill growth in STEM and non-STEM majors in life science courses of different levels at one university. Four life science classes in Fall 2022 (n = 109 students) voluntarily participated in a self-guided pollinator training module. Motivations, previous awareness, participation, and self-efficacy and self-identification for citizen science participation and for general scientific inquiry were assessed through pre- and post-surveys before and after module training. Students characterized themselves as STEM or non-STEM majors to understand self-identity. In having students self-report their identity in STEM, we found a trend (79.2%) of natural resource and agricultural majors ranking themselves as non-STEM. Across all participants, we observed a significant increase for learning outcomes between pre- and post-survey results (ɑ = 0.05). Self-reported non-STEM students showed a positive trend between surveys across survey questions. In comparison, self-reported STEM students showed very little increase across surveys but ranked highly in both pre- and post-survey results (mean = 3.42 out of 4). Overall, our findings suggest that even small-scale citizen science–based projects may increase students’ familiarity with concepts based in scientific inquiry and meet learning outcomes benefitting the goals of both higher education and citizen science initiatives.

Learning processes by exploiting restricted domain knowledge is an important task across a plethora of scientific areas, with more and more hybrid training methods additively combining data-driven and model-based approaches. Although the obtained models are more accurate than purely data-driven models, the optimization process usually comes with sensitive regularization constraints. Furthermore, while such hybrid methods have been tested in various scientific applications, they have been mostly tested on dynamical systems, with only limited study about the influence of each model component on global performance and parameter identification. In this work, we introduce a new hybrid training approach based on partial dependence, which removes the need for intricate regularization. Moreover, we assess the performance of hybrid modeling against traditional machine learning methods on standard regression problems. We compare, on both synthetic and real regression problems, several approaches for training such hybrid models. We focus on hybrid methods that additively combine a parametric term with a machine learning term and investigate model-agnostic training procedures. Therefore, experiments are carried out with different types of machine learning models, including tree-based models and artificial neural networks. We also extend our partial dependence optimization process for dynamical systems forecasting and compare it to existing schemes.

As Deep Neural Networks (DNNs) are being increasingly employed to make important simulations in rainfall‐runoff contexts, the demand for interpretability is increasing in the hydrology community. Interpretability is not just a scientific question, but rather knowing where the models fall flat, how to fix them, and how to explain their outcomes to scientific communities so that everyone understands how the model arrives at specific simulations This paper addresses these challenges by deciphering interpretable probabilistic DNNs utilizing the Deep Autoregressive Recurrent (DeepAR) and Temporal Fusion Transformer (TFT) for daily streamflow simulation across the continental United States (CONUS). We benchmarked TFT and DeepAR against conceptual to physics‐based hydrologic models. In this setting, catchment physical attributes were incorporated into the training process to create physics‐guided TFT and DeepAR configurations. Our proposed physics‐guided configurations are also designed to aggregate the patterns across the entire data set, analyze the sensitivity of key catchment physical attributes and facilitate the interpretability of temporal dynamics in rainfall‐runoff generation mechanisms. To assess the uncertainty, the modeling configurations were coupled with a quantile regression by adding Gaussian noise N(0,σ) $N\\,(0,\\sigma )$ with increasing standard deviation to the individual catchment attributes. Analysis suggested that the physics‐guided TFT was superior in predicting daily streamflow compared to the original TFT and DeepAR as well as benchmark hydrologic models. Predictive uncertainty intervals effectively bracketed most of the observational data by simultaneous simulation of various percentiles (e.g., 10th, 50th, and 90th). Interpretable physics‐guided TFT proved to be a strong candidate for CONUS daily streamflow simulations.

Hybrid Knowledge‐Guided Machine Learning (KGML) models, which are deep learning models that utilize scientific theory and process‐based model simulations, have shown improved performance over their process‐based counterparts for the simulation of water temperature and hydrodynamics. We highlight the modular compositional learning (MCL) methodology as a novel design choice for the development of hybrid KGML models in which the model is decomposed into modular sub‐components that can be process‐based models and/or deep learning models. We develop a hybrid MCL model that integrates a deep learning model into a modularized, process‐based model. To achieve this, we first train individual deep learning models with the output of the process‐based models. In a second step, we fine‐tune one deep learning model with observed field data. In this study, we replaced process‐based calculations of vertical diffusive transport with deep learning. Finally, this fine‐tuned deep learning model is integrated into the process‐based model, creating the hybrid MCL model with improved overall projections for water temperature dynamics compared to the original process‐based model. We further compare the performance of the hybrid MCL model with the process‐based model and two alternative deep learning models and highlight how the hybrid MCL model has the best performance for projecting water temperature, Schmidt stability, buoyancy frequency, and depths of different isotherms. Modular compositional learning can be applied to existing modularized, process‐based model structures to make the projections more robust and improve model performance by letting deep learning estimate uncertain process calculations. Plain Language Summary Lake models based on physical processes are powerful tools for investigating how lakes and reservoirs respond to local weather and for projecting lake responses to long‐term climate change. Historically, physical processes are the basis for designing these models. Due to an abundance of long‐term and high‐frequency data, deep learning models are used more frequently, although they do not reflect our domain expertise about hydrodynamics and heat transport. Recently, the modeling community has been focusing on merging models based on physical processes with deep learning. We are highlighting a novel methodology, modular compositional learning (MCL), that merges different modeling types in a modularized framework. Our resulting hybrid model outperformed the original model based on physical processes as well as alternative deep learning models regarding the simulation of various lake variables related to water temperature, and showed physically valid results. We are further showing various ways on how MCL can improve future lake model development and applications. Key Points Deep learning models were pretrained on process‐based lake water temperature model output and fine‐tuned on observed high‐frequency data Fine‐tuned deep learning model was integrated into process‐based model creating the hybrid model Hybrid model outperformed process‐based model and two alternative deep learning models in projecting hydrodynamic lake characteristics

Accurate aerosol optical depth (AOD) prediction remains challenging due to complex aerosol‐radiation interactions and highly variable spatio‐temporal patterns. Three critical scientific issues motivate this work: understanding whether and how physical principles can enhance deep learning predictions, identifying which aerosol properties most strongly govern AOD variations, and improving the prediction of extreme AOD events critical for air quality management. Herein, utilizing MERRA‐2 reanalysis data (1980–2024) over the Huaihe River Basin in eastern China, a Physics‐Guided deep learning framework is presented for Aerosol Optical Depth (AOD) prediction. The model proposed integrates Convolutional Neural Networks (CNN), Long Short‐TermMemory (LSTM) networks, and multi‐head attention mechanisms to capture both spatio‐temporal features and physical relationships of aerosol properties. Three key aspects are involved: First, a hybrid deep learning model is developed and evaluated, which combines CNNs for spatial correlation extraction, bidirectional LSTM for temporal dependency modeling, and multi‐head attention for feature interaction learning. Second, a comprehensive feature importance analysis is conducted by examining the relationships between different aerosol properties (mass concentration, scattering coefficient, and Ångström exponent) and AOD prediction, offering physical insights into the model's decision‐making process. Third, a specialized approach is proposed for extreme AOD event prediction, focusing on early detection and accurate forecasting of high‐AOD episodes. Overall, the results demonstrate the model's efficacy in capturing both regular AOD variations and extreme events, with the Physics‐Guided architecture showing superior performance compared to traditional methods. This integrated approach enhances AOD prediction accuracy and deepens insights into aerosol‐radiation interactions, thereby improving atmospheric monitoring and air quality forecasting. While MERRA‐2 has inherent temporal delays, this framework provides valuable capabilities for historical trend analysis, numerical model validation, and can be readily adapted for real‐time applications through transfer learning with satellite observations. Plain Language Summary Aerosol Optical Depth (AOD) is a crucial measure of how much sunlight is blocked by particles in the air, affecting both climate and air quality. Traditional methods for predicting AOD often struggle with accuracy and efficiency. This study develops a new artificial intelligence model that combines physical principles with deep learning techniques to predict AOD over the Huaihe River Basin in eastern China. Our model shows significant improvements in prediction accuracy, particularly in identifying extreme pollution events. By analyzing different types of aerosol properties, we found that the way particles scatter light is more important for predictions than their mass. The model performs better in summer than in winter, likely due to winter's more complex weather conditions. This improved prediction system could help better forecast air quality and provide early warnings for severe pollution events, benefiting public health and environmental management. Key Points A hybrid deep learning model is developed and evaluated, for feature interaction learning A comprehensive feature importance analysis is conducted by examining the relationships between different aerosol properties A specialized approach is proposed for extreme AOD event prediction

In continuous kraft pulping, key quality indicators such as Kappa number and pulp viscosity are usually measured offline at low frequency, which limits real-time quality monitoring and control. Although data-driven soft sensors have shown potential for quality prediction, their performance is often restricted by limited labeled data and weak physical consistency. In addition, existing studies have focused mainly on single-target prediction, while the coupled prediction of Kappa number and pulp viscosity remains insufficiently explored despite their common dependence on cooking conditions and degradation kinetics. To address these issues, this study proposes a physics-guided multi-task learning framework (PG-MTL) for simultaneous prediction of Kappa number and pulp viscosity. The model combines a hard-parameter-sharing multi-task architecture with a physics-guided monotonicity constraint that enforces the expected non-increasing Kappa trend with increasing H-factor. Homoscedastic uncertainty weighting is also used to balance the two regression tasks during optimization. Experiments on industrial operating data show that PG-MTL achieved R2 = 0.920 for the Kappa number and R2 = 0.910 for pulp viscosity. Compared with the strongest benchmark model, RMSE was reduced by 23.2% and 29.5% for the Kappa number and pulp viscosity, respectively. These results demonstrate that PG-MTL provides an effective and physically consistent solution for pulp-quality soft sensing under small-sample industrial conditions.

Side-scan sonar (SSS) serves as the primary perceptual instrument for Autonomous Underwater Vehicles (AUVs) in large-scale marine search and rescue (SAR) operations. However, the detection of critical targets is frequently hindered by severe hydro-acoustic noise, the spatial discontinuity of wreckage, and the weak visual signatures of small targets. To surmount these challenges, this paper presents WPG-DetNet. First, we introduce a Wavelet-Embedded Residual Backbone (WERB) to reconstruct the conventional downsampling paradigm. By substituting standard pooling with the Discrete Wavelet Transform (DWT), this architecture explicitly disentangles high-frequency noise from structural information in the frequency domain, thereby achieving the adaptive preservation of edge fidelity for large human-made targets while filtering out speckle interference. Then, addressing the distinct challenge of discontinuous aircraft wreckage, the framework further incorporates a Debris Graph Reasoning Module (D-GRM). This module models scattered fragments as nodes in a topological graph to capture long-range semantic dependencies, transforming isolated instance recognition into context-aware scene understanding. Finally, to bridge the gap between AI and underwater physics, we design a Shadow-Aided Decoupling Head (SADH) equipped with a physics-informed geometric loss. By enforcing mathematical consistency between target height and acoustic shadow length, this mechanism establishes a rigorous discriminative criterion capable of distinguishing weak-echo human bodies from seabed rocks based on shadow geometry. Experiments on the SCTD dataset demonstrate that WPG-DetNet achieves a mean Average Precision (mAP50) of 97.5% and a Recall of 96.9%. Quantitative analysis reveals that our framework outperforms the classic Faster R-CNN by a margin of 12.8% in mAP50 and surpasses the Transformer-based RT-DETR-R18 by 5.6% in high-precision localization metrics (mAP50:95). Simultaneously, WPG-DetNet maintains superior efficiency with an inference speed of 62.5 FPS and a lightweight parameter count of 16.8 M, striking an optimal balance between robust perception and the real-time constraints of AUV operations.

Deflagration fracturing is a gas-dominated, water-free reservoir stimulation technology that has shown strong potential in unconventional, low-permeability, or water-sensitive reservoirs such as coalbed methane and shale gas formations. Accurate prediction of fluid pressure variations, critical for optimizing fracture propagation and stimulation performance, is challenging. While field experiments and numerical simulations offer reliable predictions, they are hindered by high risks, costs, and computational complexity due to multi-physics coupling, Moreover, purely data-driven machine learning methods often exhibit poor generalization and may produce predictions that deviate from fundamental physical principles. To address these challenges, a physics-guided graph neural network (PG-GNN) is proposed in this study to predict the evolution of fluid pressure, the key driving factor governing fracture propagation, from a mechanistic perspective. The proposed method integrates governing equations and physical constraints to construct geometric, physical, and hybrid features and employs a graph neural network encoder to capture the spatial correlations among these features, thereby forming a deep learning framework with strong physical consistency. A multi-task loss function is further employed to balance predictive accuracy and physical rationality. Finally, the proposed model is validated using a high-resolution dataset generated by a CDEM-based numerical simulator, achieving a minimum MAPE of 0.313% and a minimum MSE of 2.309 × 10−4 on the test dataset, outperforming baseline models in both accuracy and stability and demonstrating strong extrapolation capability.

The plentiful labeled data is indispensable for data-driven intelligent fault diagnosis of rolling bearings. But in the real world, it is difficult to gather sufficient vibration signals in advance when faults occur. Selecting an intelligent model trained by other datasets to diagnose the target signals is an effective strategy in response to the data scarcity. In this paper, a guided deep subdomain adaptation network (GDSAN) is proposed to align the feature distributions across different datasets efficiently by minimizing the discrepancy between the distributions of relevant subdomains. Specifically, the proposed method realizes alignment by comparing the consistency of source labels and target pseudo labels predicted by the source classifier. The guided learning reduces the misjudgment of target pseudo labels, which helps the subdomain with identical label finding the proper common subspace more accurately. To evaluate the superiority of the proposed model, this paper conducts transfer experiments on six rolling bearing datasets and selects four mainstream deep transfer learning networks to compare with GDSAN. The results show the fault recognition accuracy of GDSAN is prominently higher than other approaches, meanwhile verify the need of using guided subdomain adaptation.