Deep Learning Techniques Research Articles

An overview of current developments and methods for identifying diabetic foot ulcers: A survey

AbstractDiabetic foot ulcers (DFUs) present a substantial health risk across diverse age groups, creating challenges for healthcare professionals in the accurate classification and grading. DFU plays a crucial role in automated health monitoring and diagnosis systems, where the integration of medical imaging, computer vision, statistical analysis, and gait information is essential for comprehensive understanding and effective management. Diagnosing DFU is imperative, as it plays a major role in the processes of diagnosis, treatment planning, and neuropathy research within automated health monitoring and diagnosis systems. To address this, various machine learning and deep learning‐based methodologies have emerged in the literature to support healthcare practitioners in achieving improved diagnostic analyses for DFU. This survey paper investigates various diagnostic methodologies for DFU, spanning traditional statistical approaches to cutting‐edge deep learning techniques. It systematically reviews key stages involved in diabetic foot ulcer classification (DFUC) methods, including preprocessing, feature extraction, and classification, explaining their benefits and drawbacks. The investigation extends to exploring state‐of‐the‐art convolutional neural network models tailored for DFUC, involving extensive experiments with data augmentation and transfer learning methods. The overview also outlines datasets commonly employed for evaluating DFUC methodologies. Recognizing that neuropathy and reduced blood flow in the lower limbs might be caused by atherosclerotic blood vessels, this paper provides recommendations to researchers and practitioners involved in routine medical therapy to prevent substantial complications. Apart from reviewing prior literature, this survey aims to influence the future of DFU diagnostics by outlining prospective research directions, particularly in the domains of personalized and intelligent healthcare. Finally, this overview is to contribute to the continual evolution of DFU diagnosis in order to provide more effective and customized medical care.This article is categorized under: Application Areas > Health Care Technologies > Machine Learning Technologies > Artificial Intelligence

Classification of Coral Reef Species using Computer Vision and Deep Learning Techniques

Classification of Coral Reef Species using Computer Vision and Deep Learning Techniques

Autonomous Scanning Tunneling Microscopy Imaging via Deep Learning.

Scanning tunneling microscopy (STM) is a powerful technique that provides the ability to manipulate and characterize individual atoms and molecules with atomic-level precision. However, the processes of scanning samples, operating the probe, and analyzing data are typically labor-intensive and subjective. Deep learning (DL) techniques have shown immense potential in automating complex tasks and solving high-dimensional problems. In this study, we developed an autonomous STM framework powered by DL to enable autonomous operations of the STM without human interventions. Our framework employs a convolutional neural network (CNN) for real-time evaluation of STM image quality, a U-net model for identifying bare surfaces, and a deep Q-learning network (DQN) agent for autonomous probe conditioning. Additionally, we integrated an object recognition model for the automated recognition of different adsorbates. This autonomous framework enables the acquisition of space-averaging information using STM techniques without compromising the high-resolution molecular imaging. We achieved measuring an area of approximately 1.9 μm2 within 48 h of continuous measurement and automatedly generated the statistics on the molecular species present within the mesoscopic area. We demonstrate the high robustness of the framework by conducting measurements at the liquid nitrogen temperature (∼78 K). We envision that the integration of DL techniques and high-resolution microscopy will not only extend the functionality and capability of scanning probe microscopes but also accelerate the understanding and discovery of new materials.

Runtime integration of machine learning and simulation for business processes: Time and decision mining predictions

Recent research in Computer Science has investigated the use of Deep Learning (DL) techniques to complement outcomes or decisions within a Discrete Event Simulation (DES) model. The main idea of this combination is to maintain a white box simulation model complement it with information provided by DL models to overcome the unrealistic or oversimplified assumptions of traditional DESs. State-of-the-art techniques in BPM combine Deep Learning and Discrete Event Simulation in a post-integration fashion: first an entire simulation is performed, and then a DL model is used to add waiting times and processing times to the events produced by the simulation model.In this paper, we aim at taking a step further by introducing Rims (Runtime Integration of Machine Learning and Simulation). Instead of complementing the outcome of a complete simulation with the results of predictions a posteriori, Rims provides a tight integration of the predictions of the DL model at runtime during the simulation. This runtime-integration enables us to fully exploit the specific predictions while respecting simulation execution, thus enhancing the performance of the overall system both w.r.t. the single techniques (Business Process Simulation and DL) separately and the post-integration approach. In particular, the runtime integration ensures the accuracy of intercase features for time prediction, such as the number of ongoing traces at a given time, by calculating them during directly the simulation, where all traces are executed in parallel. Additionally, it allows for the incorporation of online queue information in the DL model and enables the integration of other predictive models into the simulator to enhance decision point management within the process model. These enhancements improve the performance of Rims in accurately simulating the real process in terms of control flow, as well as in terms of time and congestion dimensions. Especially in process scenarios with significant congestion – when a limited availability of resources leads to significant event queues for their allocation – the ability of Rims to use queue features to predict waiting times allows it to surpass the state-of-the-art. We evaluated our approach with real-world and synthetic event logs, using various metrics to assess the simulation model’s quality in terms of control-flow, time, and congestion dimensions.

A facial and vocal emotion recognition system based on deep learning

The traditional teaching mode, mainly characterized by "cramming" instruction, often overlooks the absorption of knowledge by students during the teaching process. Even in recent years, the emerging "flipped classroom" model heavily relies on teachers' experience and expertise, making it challenging to quantify students' classroom learning states. Among young teachers lacking teaching experience, the replication of excellent teaching models faces difficulties. With the rapid development of artificial intelligence technology, deep learning and other AI techniques are increasingly applied to people's lives. Additionally, emotion computing technology is also becoming more sophisticated. In the field of education, the use of AI technology holds the potential to help teachers systematically and comprehensively assess teaching quality, promoting the sharing of high-quality educational resources nationally and globally. Existing emotion detection technologies often focus on single techniques such as image recognition or natural language processing, resulting in low accuracy in discerning human emotions and making it difficult to accurately identify emotional changes in complex scenarios such as classrooms. Furthermore, there is a lack of comprehensive emotional statistical analysis methods for audio and visual data. In light of these, we have designed and implemented a multi-modal emotion recognition and analysis system. Through testing in classroom settings, this system has the potential to assist classroom teaching and enhance teaching quality.

Arctic climate response to European radiative forcing: a deep learning study on circulation pattern changes

Abstract. Heterogeneous radiative forcing in mid-latitudes, such as that exerted by aerosols, has been found to affect the Arctic climate, though the mechanisms remain debated. In this study, we leverage deep learning (DL) techniques to explore the complex response of the Arctic climate system to local radiative forcing over Europe. We conducted sensitivity experiments using the Max Planck Institute Earth System Model (MPI-ESM1.2) coupled with atmosphere–ocean–land-surface components. Large-scale circulation patterns can mediate the impact of the forcing on Arctic climate dynamics. We employed a DL-based clustering approach to classify large-scale atmospheric circulation patterns. To enhance the analysis of how these patterns impact the Arctic climate, the poleward moist static energy transport (PMSET) associated with the atmospheric circulation patterns was incorporated as an additional similarity metric in the clustering process. Furthermore, we developed a novel method to analyze the circulation patterns' contributions to various climatic parameter anomalies. Our findings indicate that the negative radiative forcing over Europe alters existing circulation patterns and their occurrence frequency without introducing new ones. Specifically, our analysis revealed that while the regional radiative forcing alters the occurrence frequencies of the circulation patterns, these changes are not the primary drivers of the forcing's impact on the Arctic parameters. Instead, it is the shifts in the mean spatial characteristics of the atmospheric circulation patterns, induced by the forcing, that predominantly determine the effects on the Arctic climate. Our methodology facilitates the uncovering of complex, nonlinear interactions within the climate system, capturing nuances that are often obscured in broader seasonal anomaly analyses. This approach enables a deeper understanding of the dynamics driving observed climatic anomalies and their links to specific atmospheric circulation patterns.

Denoising astronomical images using an enhanced Pix2Pix model

Astronomical images are frequently affected by sensor noise, which can negatively impact the accuracy of subsequent data analysis. To address this issue, this study proposes an enhanced Pix2Pix generative adversarial network model that incorporates Residual Blocks and Self-Attention mechanisms to improve denoising performance. The effectiveness of the proposed model is evaluated by comparing it with traditional denoising methods, standard Pix2Pix, Pix2Pix with Residual Blocks, and Pix2Pix with Self-Attention using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) metrics. The findings demonstrate that the Pix2Pix model, when combined with both Residual Blocks and Self-Attention, significantly outperforms other models in noise reduction and detail preservation. This improved approach offers a robust solution for high-quality processing of astronomical images, providing clearer and more reliable data for scientific analysis. The results highlight the potential of advanced deep learning techniques in overcoming the challenges posed by sensor noise in astronomical imaging.

Facia-fix: mobile application for bell’s palsy diagnosis and assessment using computer vision and deep learning

Facial paralysis (FP) is a condition characterized by the inability to move some or all of the muscles on one or both sides of the face. Diagnosing FP presents challenges due to the limitations of traditional methods, which are time-consuming, uncomfortable for patients, and require specialized clinicians. Additionally, more advanced tools are often uncommonly available to all healthcare providers. Early and accurate detection of FP is crucial, as timely intervention can prevent long-term complications and improve patient outcomes. To address these challenges, our research introduces Facia-Fix, a mobile application for Bell's palsy diagnosis, integrating computer vision and deep learning techniques to provide real-time analysis of facial landmarks. The classification algorithms are trained on the publicly available YouTube FP (YFP) dataset, which is labeled using the House-Brackmann (HB) method, a standardized system for assessing the severity of FP. Different deep learning models were employed to classify the FP severity, such as MobileNet, CNN, MLP, VGG16, and Vision Transformer. The MobileNet model which uses transfer learning, achieved the highest performance (Accuracy: 0.9812, Precision: 0.9753, Recall: 0.9727, F1 Score: 0.974), establishing it as the optimal choice among the evaluated models. The innovation of this approach lies in its use of advanced deep learning models to provide accurate, objective, non-invasive and real-time comprehensive quantitative assessment of FP severity. Preliminary results highlight the potential of Facia-Fix to significantly improve the diagnostic and follow-up experiences for both clinicians and patients.

MRI-Based Brain Tumour Classification Using Convolutional Neural Networks: A Systematic Review and Meta-Analysis

This research assessed advancements in brain tumour classification using convolutional neural networks (CNNs) and MRI data. An analysis of 37 studies highlighted the effectiveness of CNN architectures and pre-processing methods in accurately categorising brain tumours. Issues such as class disparities and model interpretability were identified, prompting recommendations for advanced deep learning techniques, ensemble methods, and diverse datasets to enhance diagnostic accuracy. The findings underscored the importance of these methods in achieving high accuracy, with a maximum rate of 98.80% from 154 MRI images. This systematic study also included a meta-analysis from 2018 to 2022, revealing patterns in MRI cases across demographics and providing insights into healthcare trends.

A FRAMEWORK TO ANALYZE OS SYSTEMS ARTIFACTS FROM LINUX MACHINES

Memory forensics plays a pivotal role in digital investigations, providing crucial insights into the activities and artifacts of an operating system. This paper explores the application of deep learning techniques in the domain of memory forensics within the Linux environment. Linux-based systems are widely used in various contexts, including servers, embedded devices, and desktops, making memory analysis in this ecosystem of paramount importance. Traditional memory forensics techniques have relied on manual analysis, which is often time-consuming and error prone. Deep learning, a subfield of machine learning, has demonstrated remarkable capabilities in pattern recognition and feature extraction tasks. In response, this paper presents a novel framework that automates and improves memory analysis through deep learning. Key components of this framework include data collection, preprocessing, feature extraction, and model selection. We introduce a unique dataset specifically curated for Linux memory forensics, facilitating the development and evaluation of deep learning models. Our experimental results demonstrate the efficacy of using a ResNet-50 model for detecting and classifying malware from memory dumps, achieving a detection rate of 98.75% and an accuracy rate of 89% in classifying malware types. Additionally, we acknowledged the challenges and limitations of applying deep learning in memory forensics, such as model interpretability and data privacy concerns. Future research directions are discussed, including real-time memory analysis integration and techniques for handling encrypted and compressed memory data.

Evaluating Explainable Artificial Intelligence (XAI) techniques in chest radiology imaging through a human-centered Lens.

The field of radiology imaging has experienced a remarkable increase in using of deep learning (DL) algorithms to support diagnostic and treatment decisions. This rise has led to the development of Explainable AI (XAI) system to improve the transparency and trust of complex DL methods. However, XAI systems face challenges in gaining acceptance within the healthcare sector, mainly due to technical hurdles in utilizing these systems in practice and the lack of human-centered evaluation/validation. In this study, we focus on visual XAI systems applied to DL-enabled diagnostic system in chest radiography. In particular, we conduct a user study to evaluate two prominent visual XAI techniques from the human perspective. To this end, we created two clinical scenarios for diagnosing pneumonia and COVID-19 using DL techniques applied to chest X-ray and CT scans. The achieved accuracy rates were 90% for pneumonia and 98% for COVID-19. Subsequently, we employed two well-known XAI methods, Grad-CAM (Gradient-weighted Class Activation Mapping) and LIME (Local Interpretable Model-agnostic Explanations), to generate visual explanations elucidating the AI decision-making process. The visual explainability results were shared through a user study, undergoing evaluation by medical professionals in terms of clinical relevance, coherency, and user trust. In general, participants expressed a positive perception of the use of XAI systems in chest radiography. However, there was a noticeable lack of awareness regarding their value and practical aspects. Regarding preferences, Grad-CAM showed superior performance over LIME in terms of coherency and trust, although concerns were raised about its clinical usability. Our findings highlight key user-driven explainability requirements, emphasizing the importance of multi-modal explainability and the necessity to increase awareness of XAI systems among medical practitioners. Inclusive design was also identified as a crucial need to ensure better alignment of these systems with user needs.

ZeroD-fender: A Resource-aware IoT Malware Detection Engine via Fine-grained Side-channel Analysis

In early 2023, cyberattacks experienced a significant rise due to unknown (zero-day) malware targeting Internet of Things (IoT) devices. To tackle the challenge of zero-day detection within a highly resource-constrained IoT environment, we propose a novel design that utilizes fine-grained power side-channel analysis with deep learning techniques. Our approach introduces an innovative concept called multiscale feature extraction to identify the most representative malware features across diverse architectures, thereby enhancing deep learning based detection performance against zero-day malware. Specifically, we employ a fine-grained power side-channel analysis of more than 120,000 honeypot-collected malware files across a hierarchy of commands , functions , and modules to identify the unique zero-day malware behaviors. With these identified features to train our model, ZeroD-fender’s performance in detecting zero-day malware has significantly improved. In pursuit of on-device detection, we present a resource-aware online inference customization framework. This framework features our lightweight network, ThingNetV2, which uses specialized 1-D depthwise separable convolution paired with h-swish activation, leading to significant resource savings. By applying the fine-grained power analysis, ZeroD-fender demonstrates a detection rate of 95.88% across various architecture zero-day malware, achieving detection speeds ranging between 16.083 ms and 23.961 ms , depending on the specific scenario.

Machine vision-assisted genomic prediction and genome-wide association of spleen-related traits in large yellow croaker infected with visceral white-nodules disease

High-resolution and high-throughput genotype-to-phenotype studies in fish are rapidly advancing, driven by innovative technologies that aim to address the challenges of modern breeding models. In recent years, machine vision and deep learning techniques, particularly convolutional neural networks (CNNs), have achieved significant success in image recognition and segmentation. Moreover, qualitative and quantitative analysis of disease resistance has always been a crucial field of research in genetics. This motivation has led us to investigate the potential of large yellow croaker visceral white-nodules disease (VWND) in encoding information on disease resistance for the task of accession classification. In this study, we proposed an image segmentation framework for the feature extraction of the spleen after VWND infection based on machine vision. We utilized deep CNNs and threshold segmentation for automatic feature learning and object segmentation. This approach eliminates subjectivity and enhances work efficiency compared to using hand-crafted features. Additionally, we employed spleen-related traits to conduct genome-wide association analysis (GWAS), which led to the identification of 24 significant SNPs and 10 major quantitative trait loci. The results of function enrichment analysis on candidate genes also indicated potential relationships with immune regulation mechanisms. Furthermore, we explored the use of genomic selection (GS) technology for phenotype prediction of extreme individuals, which further supports the predictability of spleen-related phenotypes for VWND resistance in large yellow croakers. Our findings demonstrate that artificial intelligence (AI)-based phenotyping approaches can deliver state-of-the-art performance for genetics research. We hope this work will provide a paradigm for applying deep learning and machine vision to phenotyping in aquaculture species.

Audio Deep Fake Detection with Sonic Sleuth Model

Information dissemination and preservation are crucial for societal progress, especially in the technological age. While technology fosters knowledge sharing, it also risks spreading misinformation. Audio deepfakes—convincingly fabricated audio created using artificial intelligence (AI)—exacerbate this issue. We present Sonic Sleuth, a novel AI model designed specifically for detecting audio deepfakes. Our approach utilizes advanced deep learning (DL) techniques, including a custom CNN model, to enhance detection accuracy in audio misinformation, with practical applications in journalism and social media. Through meticulous data preprocessing and rigorous experimentation, we achieved a remarkable 98.27% accuracy and a 0.016 equal error rate (EER) on a substantial dataset of real and synthetic audio. Additionally, Sonic Sleuth demonstrated 84.92% accuracy and a 0.085 EER on an external dataset. The novelty of this research lies in its integration of datasets that closely simulate real-world conditions, including noise and linguistic diversity, enabling the model to generalize across a wide array of audio inputs. These results underscore Sonic Sleuth’s potential as a powerful tool for combating misinformation and enhancing integrity in digital communications.

Hybrid Intelligent Anomaly Detection System Using Attention based Deep Learning Approach for Cyber Attacks Prevention

Objectives: Network Intrusion Detection System (NIDS) plays an important role in finding and preventing cyber-attacks, which helps to improve the entire security posture of an organization’s network infrastructure. The development of Deep Learning (DL) techniques possess the ability of IDS to detect attacks without delay and protects from intrusions even in real-time environment. Methods: The present study proposes an improved IDS framework called Enhanced Gated Recurrent Unit Hyper-Model combined Attention Bidirectional Long-Short Term Memory (EGHAB) approach, to effectively address the detection of attacks with maximum accuracy and minimal error rate. The proposed model is enhanced by using methods like numericalization and normalisation for pre-processing the input data, Genetic Algorithm (GA) for extracting intricate features from the input data, and the EGHAB classifier, which works on the principles of both Gated Recurrent Unit (GRU) and Bidirectional Long-Short Term Memory (BiLSTM) models along with attention mechanism. Findings: The EGHAB model efficiently learns the abnormal behaviour of network and classified the attack and non-attack data using the publicly available NSL-KDD dataset and a generated real time data set, scapy. And it achieved 99.94% accuracy over NSL KDD dataset during multiclass classification and 93.3% on real time data set with reduced Error rate. Additionally, to examine the efficacy, the proposed approach is compared with other existing methods and proved its improved performance. Novelty: A combined ensemble classifier with GRU, BiLSTM and attention mechanism is designed to predict the attacks in earlier stage rather than due over. The model achieved better accuracy and reduced false assumption using anomaly prediction mechanism. Keywords: Cyber Attack, Anomaly Detection, Deep learning, Gated Recurrent Unit, Bi­ Directional Long­-Short Term Memory, Attention mechanism

Blockchain-assisted improved interval type-2 fuzzy deep learning-based attack detection on internet of things driven consumer electronics

The Internet of Things (IoTs) revolutionizes the consumer electronics landscape by presenting a degree of personalization and interactivity that was previously unimaginable. Interconnected devices are now familiar with user characteristics, giving custom skills that improve the user's satisfaction. Still, IoT remains to transform the consumer electronics field; security in IoT becomes critical, and it is utilized by cyber attackers to pose risks to public safety, compromise data privacy, gain unauthorized access, and even disrupt operations. Robust security measures are crucial for maintaining trust in the proliferation and adoption of interconnected technologies, mitigating those risks, protecting sensitive data, and certifying the integrity of the IoT ecosystem. An intrusion detection system (IDS) is paramount in IoT security, as it dynamically monitors device behaviours and network traffic to detect and mitigate any possible cyber threats. Using machine learning (ML) methods and anomaly detection algorithms, IDS can rapidly identify abnormal activities, unauthorized access, or malicious behaviours within the IoT ecosystem, thus preserving the integrity of interconnected devices and networks, safeguarding sensitive data, and protecting against cyber-attacks. This work presents an Improved Crayfish Optimization Algorithm with Interval Type-2 Fuzzy Deep Learning (ICOA-IT2FDL) technique for Intrusion Detection on IoT infrastructure. The main intention of the ICOA-IT2FDL technique is to utilize a hyperparameter-tuned improved deep learning (DL) method for intrusion detection, thereby improving safety in the IoT infrastructure. BC technology can be used to accomplish security among consumer electronics. The ICOA-IT2FDL technique employs a linear scaling normalization (LSN) approach for data normalization. In addition, features are selected using an improved crayfish optimization algorithm (ICOA). This is followed by the ICOA-IT2FDL technique, which applies the interval type-2 fuzzy deep belief network (IT2-FDBN) model to identify intrusions. Finally, the bald eagle search (BES) model strategy improves the intrusion recognition rate. A series of investigations is accomplished to ensure the enhanced accomplishment of the ICOA-IT2FDL model. The experimentation results specified that the ICOA-IT2FDL model shows better recognition results compared to recent models.

Security Assessment Framework for DDOS Attack Detection via Deep Learning

Internet of Things (IoT) is a broad network of actual, physical objects that can exchange data and communicate with other devices and systems over the Internet. Due to the scalability of the internet, the number of IoT users is increasing, and therefore it is vital to ensure the security of user info. Attacks such as DDoS (Distributed Denial of Service) and MitM (Man-in-the-Middle) can seriously interrupt operations and threaten the confidentiality and integrity of sensitive data. In this paper, a novel Iot based Security Assessment for intrusion using Deep Learning (ISAI-DL) technique has been proposed to identify IoT device vulnerabilities such as DDoS and MitM attacks. Initially, the features are extracted from the API documents using Bag of Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF) techniques. In proposed ISAI-DL technique, it triggers or activates the vulnerabilities deep within target devices by sending API documents. The vulnerability is detected via the GloVe-CNN-BiLSTM Model; if the attack occurs in an IoT device, the output is classified as either DDoS, MitM, or normal. Evaluation metrics such as accuracy, recall, precision, time efficiency, F1 score, false alarm rate, and detection rate have been used to measure how effective the proposed ISAI-DL technique is. Based on a comparative investigation, the detection rate of the proposed ISAI-DL technique is 17.23%, 18.53%, and 3.24% greater than that of the existing P2ADF, HDA-IDS, and BT-TPF techniques respectively.

Validation and Comparison of Non-stationary Cognitive Models: A Diffusion Model Application

AbstractCognitive processes undergo various fluctuations and transient states across different temporal scales. Superstatistics are emerging as a flexible framework for incorporating such non-stationary dynamics into existing cognitive model classes. In this work, we provide the first experimental validation of superstatistics and formal comparison of four non-stationary diffusion decision models in a specifically designed perceptual decision-making task. Task difficulty and speed-accuracy trade-off were systematically manipulated to induce expected changes in model parameters. To validate our models, we assess whether the inferred parameter trajectories align with the patterns and sequences of the experimental manipulations. To address computational challenges, we present novel deep learning techniques for amortized Bayesian estimation and comparison of models with time-varying parameters. Our findings indicate that transition models incorporating both gradual and abrupt parameter shifts provide the best fit to the empirical data. Moreover, we find that the inferred parameter trajectories closely mirror the sequence of experimental manipulations. Posterior re-simulations further underscore the ability of the models to faithfully reproduce critical data patterns. Accordingly, our results suggest that the inferred non-stationary dynamics may reflect actual changes in the targeted psychological constructs. We argue that our initial experimental validation paves the way for the widespread application of superstatistics in cognitive modeling and beyond.

Research on the prediction method of wing structure noise based on the combination of conditional generative adversarial neural network and numerical methods

This paper presents a technique for predicting noise generated by airfoil structures that combines deep learning techniques with traditional numerical methods. In traditional numerical methods, accurately predicting the noise of airfoil structures requires significant computational resources, making it challenging to perform low-noise optimization design for these structures. To expedite the prediction process, this study introduces Conditional Generative Adversarial Networks (CGAN). By replacing the generator and discriminator of CGAN with traditional regression neural network models, the suitability of CGAN for regression prediction is ensured. In this study, the data computation was accelerated by expanding the kernel function in the traditional boundary element method using a Taylor series. Based on the resulting data, an alternative predictive model for wing structure noise was developed by integrating Conditional Generative Adversarial Networks (CGAN). Finally, the effectiveness and feasibility of the proposed method are demonstrated through three case studies.

GeneCompass: deciphering universal gene regulatory mechanisms with a knowledge-informed cross-species foundation model.

Deciphering universal gene regulatory mechanisms in diverse organisms holds great potential for advancing our knowledge of fundamental life processes and facilitating clinical applications. However, the traditional research paradigm primarily focuses on individual model organisms and does not integrate various cell types across species. Recent breakthroughs in single-cell sequencing and deep learning techniques present an unprecedented opportunity to address this challenge. In this study, we built an extensive dataset of over 120 million human and mouse single-cell transcriptomes. After data preprocessing, we obtained 101,768,420 single-cell transcriptomes and developed a knowledge-informed cross-species foundation model, named GeneCompass. During pre-training, GeneCompass effectively integrated four types of prior biological knowledge to enhance our understanding of gene regulatory mechanisms in a self-supervised manner. By fine-tuning for multiple downstream tasks, GeneCompass outperformed state-of-the-art models in diverse applications for a single species and unlocked new realms of cross-species biological investigations. We also employed GeneCompass to search for key factors associated with cell fate transition and showed that the predicted candidate genes could successfully induce the differentiation of human embryonic stem cells into the gonadal fate. Overall, GeneCompass demonstrates the advantages of using artificial intelligence technology to decipher universal gene regulatory mechanisms and shows tremendous potential for accelerating the discovery of critical cell fate regulators and candidate drug targets.