Machine Learning Framework Research Articles

Machine Learning-Enhanced Fabrication of Three-Dimensional Co-Pt Microstructures via Localized Electrochemical Deposition

This paper presents a novel method for fabricating three-dimensional (3D) microstructures of cobalt–platinum (Co-Pt) permanent magnets using a localized electrochemical deposition (LECD) technique. The method involves the use of an electrolyte and a micro-nozzle to control the deposition process. However, traditional methods face significant challenges in controlling the thickness and uniformity of deposition layers, particularly in the manufacturing of magnetic materials. To address these challenges, this paper proposes a method that integrates machine learning algorithms to optimize the electrochemical deposition parameters, achieving a Co:Pt atomic ratio of 50:50. This optimized ratio is crucial for enhancing the material’s magnetic properties. The Co-Pt microstructures fabricated exhibit high coercivity and remanence magnetization comparable to those of bulk Co-Pt magnets. Our machine learning framework provides a robust approach for optimizing complex material synthesis processes, enhancing control over deposition conditions, and achieving superior material properties. This method opens up new possibilities for the fabrication of 3D microstructures with complex shapes and structures, which could be useful in a variety of applications, including micro-electromechanical systems (MEMSs), micro-robots, and data storage devices.

Ensemble based high performance deep learning models for fake news detection

Social media has emerged as a dominant platform where individuals freely share opinions and communicate globally. Its role in disseminating news worldwide is significant due to its easy accessibility. However, the increase in the use of these platforms presents severe risks for potentially misleading people. Our research aims to investigate different techniques within machine learning, deep learning, and ensemble learning frameworks in Arabic fake news detection. We integrated FastText word embeddings with various machine learning and deep learning methods. We then leveraged advanced transformer-based models, including BERT, XLNet, and RoBERTa, optimizing their performance through careful hyperparameter tuning. The research methodology involves utilizing two Arabic news article datasets, AFND and ARABICFAKETWEETS datasets, categorized into fake and real subsets and applying comprehensive preprocessing techniques to the text data. Four hybrid deep learning models are presented: CNN-LSTM, RNN-CNN, RNN-LSTM, and Bi-GRU-Bi-LSTM. The Bi-GRU-Bi-LSTM model demonstrated superior performance regarding the F1 score, accuracy, and loss metrics. The precision, recall, F1 score, and accuracy of the hybrid Bi-GRU-Bi-LSTM model on the AFND Dataset are 0.97, 0.97, 0.98, and 0.98, and on the ARABICFAKETWEETS dataset are 0.98, 0.98, 0.99, and 0.99 respectively. The study’s primary conclusion is that when spotting fake news in Arabic, the Bi-GRU-Bi-LSTM model outperforms other models by a significant margin. It significantly aids the global fight against false information by setting the stage for future research to expand fake news detection to multiple languages.

A Machine Learning Framework for Classroom EEG Recording Classification: Unveiling Learning-Style Patterns

Classroom EEG recordings classification has the capacity to significantly enhance comprehension and learning by revealing complex neural patterns linked to various cognitive processes. Electroencephalography (EEG) in academic settings allows researchers to study brain activity while students are in class, revealing learning preferences. The purpose of this study was to develop a machine learning framework to automatically classify different learning-style EEG patterns in real classroom environments. Method: In this study, a set of EEG features was investigated, including statistical features, fractal dimension, higher-order spectra, entropy, and a combination of all sets. Three different machine learning classifiers, random forest (RF), K-nearest neighbor (KNN), and multilayer perceptron (MLP), were used to evaluate the performance. The proposed framework was evaluated on the real classroom EEG dataset, involving EEG recordings featuring different teaching blocks: reading, discussion, lecture, and video. Results: The findings revealed that statistical features are the most sensitive feature metric in distinguishing learning patterns from EEG. The statistical features and RF classifier method tested in this study achieved an overall best average accuracy of 78.45% when estimated by fivefold cross-validation. Conclusions: Our results suggest that EEG time domain statistics have a substantial role and are more reliable for internal state classification. This study might be used to highlight the importance of using EEG signals in the education context, opening the path for educational automation research and development.

Machine Learning Framework for Conotoxin Class and Molecular Target Prediction

Conotoxins are small and highly potent neurotoxic peptides derived from the venom of marine cone snails which have captured the interest of the scientific community due to their pharmacological potential. These toxins display significant sequence and structure diversity, which results in a wide range of specificities for several different ion channels and receptors. Despite the recognized importance of these compounds, our ability to determine their binding targets and toxicities remains a significant challenge. Predicting the target receptors of conotoxins, based solely on their amino acid sequence, remains a challenge due to the intricate relationships between structure, function, target specificity, and the significant conformational heterogeneity observed in conotoxins with the same primary sequence. We have previously demonstrated that the inclusion of post-translational modifications, collisional cross sections values, and other structural features, when added to the standard primary sequence features, improves the prediction accuracy of conotoxins against non-toxic and other toxic peptides across varied datasets and several different commonly used machine learning classifiers. Here, we present the effects of these features on conotoxin class and molecular target predictions, in particular, predicting conotoxins that bind to nicotinic acetylcholine receptors (nAChRs). We also demonstrate the use of the Synthetic Minority Oversampling Technique (SMOTE)-Tomek in balancing the datasets while simultaneously making the different classes more distinct by reducing the number of ambiguous samples which nearly overlap between the classes. In predicting the alpha, mu, and omega conotoxin classes, the SMOTE-Tomek PCA PLR model, using the combination of the SS and P feature sets establishes the best performance with an overall accuracy (OA) of 95.95%, with an average accuracy (AA) of 93.04%, and an f1 score of 0.959. Using this model, we obtained sensitivities of 98.98%, 89.66%, and 90.48% when predicting alpha, mu, and omega conotoxin classes, respectively. Similarly, in predicting conotoxins that bind to nAChRs, the SMOTE-Tomek PCA SVM model, which used the collisional cross sections (CCSs) and the P feature sets, demonstrated the highest performance with 91.3% OA, 91.32% AA, and an f1 score of 0.9131. The sensitivity when predicting conotoxins that bind to nAChRs is 91.46% with a 91.18% sensitivity when predicting conotoxins that do not bind to nAChRs.

Uncovering coordinated cross-platform information operations: Threatening the integrity of the 2024 U.S. presidential election

Information operations (IOs) pose a significant threat to the integrity of democratic processes, with the potential to influence election-related online discourse. In anticipation of the 2024 U.S. presidential election, we present a study aimed at uncovering the digital traces of coordinated IOs on X (formerly Twitter). Using our machine learning framework for detecting online coordination, we analyze a dataset comprising election-related conversations on X from May to July 2024. This reveals a network of coordinated inauthentic actors, displaying notable similarities in their link-sharing behaviors. Our analysis shows concerted efforts by these accounts to disseminate misleading, redundant, and biased information across the Web through a coordinated cross-platform information operation: The links shared by this network frequently direct users to other social media platforms or mock news sites featuring low-quality political content and, in turn, promoting the same X and YouTube accounts. Members of this network also shared deceptive images generated by AI, accompanied by language attacking political figures and symbolic imagery intended to convey power and dominance. While X has suspended or restricted a subset of these accounts, 75 percent ofthe coordinated network remains active, garnering substantial traction over time: The suspicious Web sites promoted by this coordinated network are shared thousands of times per day by the X user base, further amplifying their reach and potential impact. Our findings underscore the critical role of developing computational models to scale up the detection of threats on large social media platforms, and emphasize the broader implications of these techniques to detect IOs across the wider Web.

Machine learning powered predictive modelling of complex residual stress for nuclear fusion reactor design

Fusion In-vessel components, assembled and maintained using laser welding, one of the most promising techniques, exhibit complex distributions of residual stress, microstructures, and material properties. These residual stresses can compromise structural integrity and lifespan of critical components. Although using advanced experimental measurements can evaluate the residual stress for individual case, extending the measurements to massive number of components are costly and time-consuming. Traditional machine learning (ML) models struggle to account for the heterogeneity and anisotropy of these stress distributions. Here, we develop a novel ML framework based on the Eurofer97 steel, the structural material for in-vessel components. The ML framework is trained on high-resolution residual stress data derived from recently-developed evaluation techniques. Combining with microstructures, the model enables prediction of heterogenous and anisotropic residual stress distribution. It successfully predicts the compressive residual stress in fusion zone (∼−200 MPa) balanced by tensile residual stress in heat affected zone (∼300 MPa), aligning closely with experimental results with the R-squared value of 0.989 and the mean square error of 10−4. Unlike experiments that take hours, the ML model provides predictions within seconds. It offers valuable insights into residual stress prediction for various joints, enhancing the reliability and lifetime prediction of in-vessel components.

Thermal Stratification Prediction in Reactor System Based on CFD Simulations Accelerated by A Data-driven Coarse-grid Turbulence model

Thermal stratification in large enclosures is an integral phenomenon to nuclear reactor system safety. Currently, the effective model for thermal stratification utilizes a multi-scale method that integrates 1-D system-level and 3-D CFD code, which offers thermal stratification details while supplying system-level data across various domains. Nonetheless, harmonizing two codes that operate on different spatial and temporal scales presents a significant challenge, with high-resolution CFD simulations requiring substantial computational resources. This study introduced a data-driven coarse-grid turbulence model based on local flow characteristics at a significantly coarser scale targeting improved efficiency and accuracy in reactor safety analysis concerning thermal stratification. A machine learning framework has been introduced to expedite the RANS-solving process by coupling of OpenFOAM and TensorFlow, which entails training a deep neural network with fine-grid CFD-generated data to predict turbulent eddy viscosity. The feasibility of the developed data-driven turbulence model was proven through the SUPERCAVNA experimental facility problem validation.

HyperCAN: Hypernetwork-driven deep parameterized constitutive models for metamaterials

We introduce HyperCAN, a machine learning framework that utilizes hypernetworks to construct adaptable constitutive artificial neural networks for a wide range of beam-based metamaterials exhibiting diverse mechanical behavior under finite deformations. HyperCAN integrates an input convex neural network that models the nonlinear stress–strain map of a truss lattice, while ensuring adherence to fundamental mechanics principles, along with a hypernetwork that dynamically adjusts the parameters of the convex network as a function of the lattice topology and geometry. This unified framework demonstrates robust generalization in predicting the mechanical behavior of previously unseen metamaterial designs and loading scenarios well beyond the training domain. We show how HyperCAN can be integrated into multiscale simulations to accurately capture the highly nonlinear responses of large-scale truss metamaterials, closely matching fully resolved simulations while significantly reducing computational costs. This offers new efficient opportunities for the multiscale design and optimization of truss metamaterials.

Machine learning framework for selective and sensitive metal ion sensing with nitrogen-doped graphene quantum dots heterostructure

This study introduces a machine learning (ML) framework to optimize photodetector performance for sensor applications. Using the data from the fabricated photodetector with the heterostructure of nitrogen-doped graphene quantum dot and gold nanoparticles (Au@N-GQDs), various supervised ML models (more than 20 models) are trained and tested for the selection and refinement of the most effective algorithm for our work. Depending on the best-performed ML model, the optimized working wavelength of the photodetector is found for the detection of metal ions. Remarkably, the ML-based sensor shows a high level of selectivity and sensitivity in nM level towards Fe3+ ions in Brahmaputra river water. A strong alignment between model predictions and experimental outcomes validates the efficacy of the proposed ML-based framework. Moreover, data visualization techniques such as heatmaps, classification algorithms, and confusion matrices are introduced to identify the trends in the database. The mechanistic insight of the sensor performance towards Fe3+ ion sensing is further explained with heatmap analysis and experimental verification, which emphasizes the role of photo-induced charge transfer and Fe–O bond formation between metal ions and Au@N-GQDs due to the high electron affinity of Fe3+ ions.

Enabling scalable and adaptive machine learning training via serverless computing on public cloud

In today’s production machine learning (ML) systems, models are continuously trained, improved, and deployed. ML design and training are becoming a continuous workflow of various tasks that have dynamic resource demands. Serverless computing is an emerging cloud paradigm that provides transparent resource management and scaling for users and has the potential to revolutionize the routine of ML design and training. However, hosting modern ML workflows on existing serverless platforms has non-trivial challenges due to their intrinsic design limitations such as stateless nature, limited communication support across function instances, and limited function execution duration. These limitations result in a lack of an overarching view and adaptation mechanism for training dynamics, and an amplification of existing problems in ML workflows.To address the above challenges, we propose SMLT, an automated, scalable and adaptive serverless framework on public cloud to enable efficient and user-centric ML design and training. SMLT employs an automated and adaptive scheduling mechanism to dynamically optimize the deployment and resource scaling for ML tasks during training. SMLT further enables user-centric ML workflow execution by supporting user-specified training deadline and budget limit. In addition, by providing an end-to-end design, SMLT solves the intrinsic problems in public cloud serverless platforms such as the communication overhead, limited function execution duration, need for repeated initialization, and also provides explicit fault tolerance for ML training. SMLT is open-sourced and compatible with all major ML frameworks. Our experimental evaluation with large, sophisticated modern ML models demonstrates that SMLT outperforms the state-of-the-art VM-based systems and existing public cloud serverless ML training frameworks in both training speed (up to 8×) and monetary cost (up to 3×).

3-D full-field reconstruction of chemically reacting flow towards high-dimension conditions through machine learning

Monitoring chemically reacting flow is crucial for optimizing and controlling the conversion processes in various chemical engineering scenarios. These processes are often influenced by multiple factors, creating a high-dimensional condition space that imposes intensive computational cost to computational fluid dynamics (CFD) methods. We present a machine learning (ML) framework to reconstruct the 3-D scalar field preserving the accuracy and resolution of CFD modeling. We demonstrate the framework’s effectiveness through a case study on biomass combustion in an industrial-scale grate furnace, which exemplifies the challenges of modeling chemically reacting flow in complex geometric domains influenced non-linearly by multiple parameters. The framework implements three core principles: (1) a Reactor-Structure-Resembled Network (RSRNet) reflecting the operational logics of the reactor, (2) an incremental learning strategy based on random Latin Hypercube Sampling, and (3) the incorporation of self-extracted high-level features from data as soft constraints in the loss function. Ground truth data were calculated through a CFD model, with all cases sampled from a 11-variable condition space. Using only 240 randomly sampled cases (around 1 % total cases in the discretized condition space), RSRNet precisely replicated 1-D and 2-D scalar distributions through incremental training. Further reconstruction of 3-D temperature field only used 40 3-D CFD cases, resulting in the training and testing errors decreased to 7.94 × 10-4 and 1.55 × 10-3, respectively, approaching the level of the 2-D reconstruction, with the average error in temperature field predictions not exceeding 50 K. The generalization ability was validated using a separate test dataset with continuously changing excess air ratio and capacity ratio in wide ranges, including some extreme values, and the model successfully captured complex phenomena under unseen conditions.

Multiparametric Characterization of Focal Cortical Dysplasia Using 3D MR Fingerprinting.

To develop a multiparametric machine-learning (ML) framework using high-resolution 3 dimensional (3D) magnetic resonance (MR) fingerprinting (MRF) data for quantitative characterization of focal cortical dysplasia (FCD). We included 119 subjects, 33 patients with focal epilepsy and histopathologically confirmed FCD, 60 age- and gender-matched healthy controls (HCs), and 26 disease controls (DCs). Subjects underwent whole-brain 3 Tesla MRF acquisition, the reconstruction of which generated T1 and T2 relaxometry maps. A 3D region of interest was manually created for each lesion, and z-score normalization using HC data was performed. We conducted 2D classification with ensemble models using MRF T1 and T2 mean and standard deviation from gray matter and white matter for FCD versus controls. Subtype classification additionally incorporated entropy and uniformity of MRF metrics, as well as morphometric features from the morphometric analysis program (MAP). We translated 2D results to individual probabilities using the percentage of slices above an adaptive threshold. These probabilities and clinical variables were input into a support vector machine for individual-level classification. Fivefold cross-validation was performed and performance metrics were reported using receiver-operating-characteristic-curve analyses. FCD versus HC classification yielded mean sensitivity, specificity, and accuracy of 0.945, 0.980, and 0.962, respectively; FCD versus DC classification achieved 0.918, 0.965, and 0.939. In comparison, visual review of the clinical magnetic resonance imaging (MRI) detected 48% (16/33) of the lesions by official radiology report. In the subgroup where both clinical MRI and MAP were negative, the MRF-ML models correctly distinguished FCD patients from HCs and DCs in 98.3% of cross-validation trials. Type II versus non-type-II classification exhibited mean sensitivity, specificity, and accuracy of 0.835, 0.823, and 0.83, respectively; type IIa versus IIb classification showed 0.85, 0.9, and 0.87. In comparison, the transmantle sign was present in 58% (7/12) of the IIb cases. The MRF-ML framework presented in this study demonstrated strong efficacy in noninvasively classifying FCD from normal cortex and distinguishing FCD subtypes. ANN NEUROL 2024;96:944-957.

Assessing spatiotemporal variations of soil organic carbon and its vulnerability to climate change: A bottom-up machine learning approach

Soil organic carbon (SOC) is a crucial component of the terrestrial carbon cycle and essential for agricultural productivity. Quantifying its sensitivity to future climate change is vital for sustaining agricultural practices and mitigating greenhouse gas emissions. However, this remains a challenge as long-term SOC data are scarce and substantial uncertainties regarding future climate scenarios. This study presents a bottom-up machine learning framework to assess the spatiotemporal variations of SOC and its vulnerability to climate change in the Jinghe River Basin, a typical loess hilly and gully watershed. Firstly, the long-term (2000–2023) dynamics of SOC was estimated by integrating in-situ measurements with machine learning techniques. Results show that the high SOC values are primarily distributed in the farmland of the mountain-loess transition zone, while the low-value areas are mainly found in the loess region. During the study period, the SOC content exhibited a slight increasing trend with a rate of 0.02 ​g ​kg−1 ​yr−1 (p ​= ​0.449). The vulnerability of farmland surface SOC to future climate change was then evaluated by combining a robust machine learning model with the bottom-up framework. To this end, the study explored a wide range of possible future climates to identify critical climate thresholds and their spatial variation across the basin’s farmlands. Based on this analysis, this research found that the farmland in the northern basin is generally more susceptible to changing climate with even marginal rises in temperature could lead to severe loss in SOC. These results highlight the need for proactive climate adaptation strategies to safeguard SOC in vulnerable agricultural landscapes, ensuring soil health and resilience in the face of climate change.

Reconstruction of PM2.5 Concentrations in East Asia on the Basis of a Wide–Deep Ensemble Machine Learning Framework and Estimation of the Potential Exposure Level from 1981 to 2020

Satellite observations are widely used to estimate the concentrations of surface air pollutants, but the temporal coverage of these datasets is relatively short. To overcome this limitation, we propose a wide–deep ensemble machine learning framework to reconstruct the fine particulate matter (PM2.5) dataset of east Asia (EA) over the past four decades (1981–2020). The results indicate that the framework effectively leveraged the advantages of satellite observations (higher accuracy) and model-based estimations (longer temporal coverage) of surface air pollutants. The reconstructed PM2.5 concentrations agreed well with the ground measurements, with coefficient of determination (R2) and root-mean-square error (RMSE) values of 0.99 and 1.38 μg·m−3, respectively, which outperformed the satellite-based PM2.5 estimates. As more ground measurements were incorporated into the model for training, the average RMSE in Japan and the Korean Peninsula decreased to 0.83 and 1.50 μg·m−3, respectively. Simultaneously, on the basis of the reconstructed datasets, we investigated the exposure level to PM2.5 in EA from 1981 to 2020. Since 2000, the increase in anthropogenic emissions has substantially worsened the air quality in EA, and nearly 50% of the population resided in areas where the annual average PM2.5 concentrations exceeded 50 μg·m−3 from 2009–2010. Despite the implementation of various mitigation strategies by local authorities to lower the ambient PM2.5 concentrations, the entire exposure level in EA is still implausible to meet the WHO air quality guidelines. In addition, population aging and climate change have the potential to increase PM2.5 exposure risk in the future. For policy-makers in EA, it is essential to consider the effects of these factors and develop more effective mitigation strategies that aim to lessen the health impact associated with PM2.5 exposure.

A machine learning framework for seismic risk assessment of industrial equipment

The paper aims to propose a novel machine learning framework for seismic risk assessment of industrial facilities. In this respect, a compound artificial neural network model is employed, which is based on two different artificial neural network models in series. The first artificial neural network is a regression model employed to generate samples of a vector-valued intensity measure. The second one is a classification model that is used to predict structural damage, starting from the outcomes of the first artificial neural network model. The datasets used for training and validation of the two artificial neural networks are based on hazard-consistent accelerograms and numerical analyses that are performed with an efficient finite element model of the structure. The methodology entails a preliminary feature selection phase for the identification of the aforementioned vector-valued of intensity measures that better classifies the damage/no-damage condition of the structure. This phase is implemented through the principal component analysis method. Subsequently, the Metropolis–Hastings algorithm is used to generate samples of a selected intensity measure, feeding the first ANN model. In turn, the chosen features are used as input parameters of the second ANN model to generate samples of damage/no-damage events. Using the two ANN in series, the mean annual frequency of exceeding a specific limit state is derived. The proposed framework is validated using a typical multi-storey steel frame, focusing on the seismic risk assessment of a vertical storage tank located at the first floor of the primary structure. The proposed method exhibits some clear advantages of combining numerical models with ANN techniques, mainly related to: a reduced computational time; the avoidance of any prior information on the probabilistic model of fragility curves; and the use of model-driven data.

DeepForest-HTP: A Novel Deep Forest Approach for Predicting Antihypertensive Peptides

Hypertension is a major preventable risk factor for cardiovascular disease, affecting over 1.5 billion adults worldwide. Antihypertensive peptides (AHTPs) have gained attention as a natural therapeutic option with minimal side effects. This study proposes a Deep Forest-based machine learning framework for AHTP prediction, leveraging a multi-granularity cascade structure to enhance classification accuracy. We integrated data from BIOPEP-UWM and three previously used datasets, totaling 2000 peptide sequences, and introduced novel feature extraction methods to build a comprehensive dataset for model training.This study represents the first application of Deep Forest for AHTP identification, demonstrating substantial classification performance advantages over traditional methods (e.g., SVM, CNN, and XGBoost) as well as recent mainstream prediction models (Ensemble-AHTPpred, CNN-SVM Ensemble, and mAHTPred). Requiring no complex manual feature engineering, the model adapts flexibly to various data needs, offering a novel perspective for efficient AHTP prediction and promising utility in hypertension management.On the benchmark dataset, the model achieved high accuracy, sensitivity, and AUC, providing a robust tool for identifying safe and effective AHTPs. However, future efforts should incorporate larger and more diverse independent validation datasets to further improve the model and enhance its generalizability. Additionally, the model's predictive accuracy relies on known AHTP targets and sequence features, potentially limiting its ability to detect AHTPs with uncharacterized or atypical properties.

Functional brain changes in Mexican women with fibromyalgia

Fibromyalgia (FM) is a chronic condition marked by widespread pain, fatigue, sleep problems, cognitive decline, and other symptoms. Despite extensive research, the pathophysiology of FM remains poorly understood, complicating diagnosis and treatment, which often relies on self-report questionnaires. This study explored structural and functional brain changes in women with FM, identified potential biomarkers, and examined their relationship with FM severity. MRI data from 33 female FM patients and 33 matched healthy controls were utilized, focusing on T1-weighted MRI and resting-state fMRI scans. Functional connectivity (FC) analysis was performed using a machine learning framework to differentiate FM patients from healthy controls and predict FM symptom severity. No significant differences were found in brain structural features, such as gray matter volume, white matter volume, deformation-based morphometry, and cortical thickness. However, significant differences in FC were observed between FM patients and healthy controls, particularly in the default mode network (DMN), somatomotor network (SMN), visual network (VIS), and dorsal attention network (DAN). The FC metrics were significantly associated with FM severity. Our prediction model differentiated FM patients from healthy controls with an area under the curve of 0.65. FC measures accurately estimated FM symptom severities with a significant correlation (r = 0.45, p = 0.007). Functional connections in the DMN, VIS, and DAN were crucial in determining FM severity. These findings suggest that integrating brain FC measurements could serve as valuable biomarkers for identifying FM patients from healthy individuals and predicting FM symptom severity, improving diagnostic accuracy and facilitating the development of targeted therapeutic strategies.

Benchmarking and building DNA binding affinity models using allele-specific and allele-agnostic transcription factor binding data

BackgroundTranscription factors (TFs) bind to DNA in a highly sequence-specific manner. This specificity manifests itself in vivo as differences in TF occupancy between the two alleles at heterozygous loci. Genome-scale assays such as ChIP-seq currently are limited in their power to detect allele-specific binding (ASB) both in terms of read coverage and representation of individual variants in the cell lines used. This makes prediction of allelic differences in TF binding from sequence alone desirable, provided that the reliability of such predictions can be quantitatively assessed.ResultsWe here propose methods for benchmarking sequence-to-affinity models for TF binding in terms of their ability to predict allelic imbalances in ChIP-seq counts. We use a likelihood function based on an over-dispersed binomial distribution to aggregate evidence for allelic preference across the genome without requiring statistical significance for individual variants. This allows us to systematically compare predictive performance when multiple binding models for the same TF are available. To facilitate the de novo inference of high-quality models from paired-end in vivo binding data such as ChIP-seq, ChIP-exo, and CUT&Tag without read mapping or peak calling, we introduce an extensible reimplementation of our biophysically interpretable machine learning framework named PyProBound. Explicitly accounting for assay-specific bias in DNA fragmentation rate when training on ChIP-seq yields improved TF binding models. Moreover, we show how PyProBound can leverage our threshold-free ASB likelihood function to perform de novo motif discovery using allele-specific ChIP-seq counts.ConclusionOur work provides new strategies for predicting the functional impact of non-coding variants.

Federated Learning for Predicting Postoperative Remission of Patients with Acromegaly: A Multicentered study

BackgroundDecentralized federated learning (DFL) may serve as a useful framework for machine learning (ML) tasks in multicentered studies, maximizing the use of clinical data without data sharing. We aim to propose the first workflow of DFL for ML tasks in multicentered studies, which can be as powerful as those using centralized data. MethodsA DFL workflow was developed with four steps: registration, local computation, model update, and inspection. A total of 598 participants with acromegaly from PUMCH, and 120 participants from XWH were enrolled. The cohort from PUMCH was further split into five centers. Nine clinical features were incorporated into ML-based models trained based on four algorithms: LR, GBDT, SVM, and DNN. The area under the curve (AUC) of receiver operating characteristic curves was used to evaluate the performance of the models. ResultsModels trained based on DFL workflow performed better than most models in LR (P<0.05), all models in DNN, SVM, and GBDT (P<0.05). Models trained on DFL workflow performed as powerful as models trained on centralized data in LR, DNN, and SVM (P>0.05). ConclusionsWe demonstrate that the DFL workflow without data sharing should be a more appropriate method in ML tasks in multicentered studies. And the DFL workflow should be further exploited in clinical researches in other departments and it can encourage and facilitate multicentered studies.

Machine learning for predictive AAC: Improving speech and gesture-based communication systems

Augmentative and alternative communication (AAC) systems play a crucial role in supporting individuals with severe communication disabilities by providing accessible means of expression and engagement. However, many conventional AAC devices rely on manual input or basic predictive functions, which can limit communication efficiency and responsiveness. The application of machine learning (ML) to AAC offers new opportunities to enhance these systems, enabling them to provide faster, more accurate, and contextually relevant communication assistance. Advances in ML, particularly in predictive text, speech recognition, and gesture interpretation, allow AAC systems to adapt more intuitively to user needs, predicting intent based on usage patterns and multimodal data, such as voice and gestures. Current research highlights the potential of ML to address key gaps in AAC technology by creating more responsive, personalized systems that align with individual user behaviours. This study proposes a novel ML framework designed to integrate these capabilities, promising improvements in communication speed, user autonomy, and accuracy. By addressing the challenges and limitations of traditional AAC devices, this research aims to advance accessible communication solutions that empower users and improve quality of life.