A5059009894- Genetic Associations and Epidemiology
- Bioinformatics and Genomic Networks
- Retinal Development and Disorders
- Cell Image Analysis Techniques
- Anomaly Detection Techniques and Applications
- Single-cell and spatial transcriptomics
- Glaucoma and retinal disorders
- Gene expression and cancer classification
- Retinal Imaging and Analysis
- Cancer Genomics and Diagnostics
- Genetic Mapping and Diversity in Plants and Animals
- Neural dynamics and brain function
- COVID-19 diagnosis using AI
- Integrated Circuits and Semiconductor Failure Analysis
- Artificial Intelligence in Healthcare
- COVID-19 epidemiological studies
- Machine Learning in Healthcare
- Advanced Proteomics Techniques and Applications
- Health, Environment, Cognitive Aging
- Genomics and Rare Diseases
- Metabolomics and Mass Spectrometry Studies
- Adversarial Robustness in Machine Learning
- Sleep and Wakefulness Research
- Statistical Methods in Clinical Trials
- Functional Brain Connectivity Studies
California University of Pennsylvania
2025
Purdue University West Lafayette
2021-2024
Cardiovascular health interacts with cognitive and mental in complex ways, yet little is known about the phenotypic genetic links of heart-brain systems. We quantified connections using multiorgan magnetic resonance imaging (MRI) data from more than 40,000 subjects. Heart MRI traits displayed numerous association patterns brain gray matter morphometry, white microstructure, functional networks. identified 80 associated genomic loci (
Abstract The retina, an anatomical extension of the brain, forms physiological connections with visual cortex brain. Although retinal structures offer a unique opportunity to assess brain disorders, their relationship structure and function is not well understood. In this study, we conducted systematic cross-organ genetic architecture analysis eye-brain using imaging endophenotypes. We identified novel phenotypic links between biomarkers measures from multimodal magnetic resonance (MRI),...
Graph Anomaly Detection (GAD) is a technique used to identify abnormal nodes within graphs, finding applications in network security, fraud detection, social media spam and various other domains. A common method for GAD Auto-Encoders (GAEs), which encode graph data into node representations anomalies by assessing the reconstruction quality of graphs based on these representations. However, existing GAE models are primarily optimized direct link reconstruction, resulting connected being...
Plasma protein levels provide important insights into human disease, yet a comprehensive assessment of plasma proteomics across organs is lacking. Using large-scale multimodal data from the UK Biobank, we integrated with organ imaging to map their phenotypic and genetic links, analyzing 2,923 proteins 1,051 traits multiple organs. We uncovered 5,067 protein-imaging associations, identifying both organ-specific organ-shared proteomic relations, along enriched protein-protein interaction...
Polygenic risk scores (PRS) are becoming increasingly vital for prediction and stratification in precision medicine. However, PRS model training presents significant challenges broader adoption of PRS, including limited access to computational resources, difficulties implementing advanced methods, availability privacy concerns over individual-level genetic data. Cloud computing provides a promising solution with centralized data resources. Here we introduce PennPRS ( https://pennprs.org ),...
Abstract As an anatomical extension of the brain, retina eye is synaptically connected to visual cortex, establishing physiological connections between and brain. Despite unique opportunity retinal structures offer for assessing brain disorders, less known about their relationship structure function. Here we present a systematic cross-organ genetic architecture analysis eye-brain using imaging endophenotypes. Novel phenotypic links were identified biomarkers function measures derived from...
Recently, machine learning models have achieved tremendous success in prioritizing candidate genes for genetic diseases. These are able to accurately quantify the similarity among disease and based on intuition that similar more likely be associated with However, features these methods rely often hard collect due high experimental cost various other technical limitations. Existing solutions of this problem significantly increase risk overfitting decrease generalizability models.In work, we...
Abstract Motivation We have entered the multi-omics era and can measure cells from different aspects. Hence, we get a more comprehensive view by integrating or matching data spaces corresponding to same object. However, it is particularly challenging in single-cell scenario because such are very sparse with extremely high dimensions. Though some techniques be used scATAC-seq scRNA-seq simultaneously, usually highly noisy due limitations of experimental environment. Results To promote...
Abstract The human cerebral cortex is a vital component of brain function, but the genetic influences on cortical functional organization remain poorly understood. In this study, we used parcellation-based approach to process resting-state and task-evoked magnetic resonance imaging (fMRI) from over 48,000 individuals in UK Biobank ABCD studies. We identified 47 loci associated with areas networks at rest, 15 which also affected connectivity during task performance. observed patterns...
Sleep is crucial for overall physical and mental health, concerning organs such as the brain, heart, eye, liver, kidney, lung. Nonetheless, a thorough understanding of how sleep relates to these organs, well their genetic bases, remains elusive. Here we conducted systematic mapping sleep-organ connections, focusing on 623 multi-organ imaging biomarkers 10 traits. Both phenotypic analyses uncovered robust associations between structure function multiple brain functions measured by functional...
The UK Biobank (UKB) imaging project is a crucial resource for biomedical research, but limited to 100,000 participants due cost and accessibility barriers. Here we used genetic data predict heritable imaging-derived phenotypes (IDPs) larger cohort. We developed evaluated 4,375 IDP scores (IGS) derived from UKB brain body images. When applied who were not imaged, IGS revealed links numerous stratified at increased risk both somatic diseases. For example, identified individuals higher...
As large-scale biobanks provide increasing access to deep phenotyping and genomic data, genome-wide association studies (GWAS) are rapidly uncovering the genetic architecture behind various complex traits diseases. GWAS publications typically make their summary-level data (GWAS summary statistics) publicly available, enabling further exploration of overlaps between phenotypes gathered from different cohorts. However, systematically analyzing high-dimensional statistics for thousands can be...
ABSTRACT We have entered the multi-omics era, and we can measure cells from different aspects. When dealing with such data, first step is to determine correspondence among omics. In other words, should match data spaces corresponding same object. This problem particularly challenging in single-cell scenario because are very sparse extremely high dimensions. Secondly, matched rare hard collect. Furthermore, due limitations of experimental environment, usually highly noisy. To promote...
Abstract Understanding the complex causal relationships among major clinical outcomes and interplay multiple organs remains a significant challenge. By using imaging phenotypes, we can characterize functional structural architecture of human organs. Mendelian randomization (MR) provides valuable framework for inferring causality by leveraging genetic variants as instrumental variables. In this study, conducted systematic multi-organ MR analysis involving 402 traits 372 outcomes. Our revealed...
Abstract Brain ventricular and subcortical structures are heritable both in size shape. Genetic influences on brain region have been studied using conventional volumetric measures, but little is known about the genetic basis of shapes. Here we developed pipelines to extract seven complementary shape measures for lateral ventricles, structures, hippocampal subfields. Based over 45,000 subjects UK Biobank ABCD studies, 60 loci were identified be associated with features ( P < 1.09 × 10 -10...
Deep neural network (DNN) is a popular model implemented in many systems to handle complex tasks such as image classification, object recognition, natural language processing etc. Consequently DNN structural vulnerabilities become part of the security those systems. In this paper we study root cause adversarial examples. We examine response surface understand its classification boundary. Our reveals problem boundary that leads Existing attack algorithms can generate from handful few hundred...
Muilti-modality data are ubiquitous in biology, especially that we have entered the multi-omics era, when can measure same biological object (cell) from different aspects (omics) to provide a more comprehensive insight into cellular system. When dealing with such data, first step is determine correspondence among modalities. In other words, should match spaces corresponding object. This problem particularly challenging single-cell scenario because very sparse extremely high dimensions....