Recent advances in spatial transcriptomics (ST) and cost reductions have enabled large-scale multi-slice ST data generation, enhancing the statistical power to detect subtle biological signals. However, cross-slice inconsistencies and data quality variability present significant analytical challenges. To overcome these limitations, we developed MUSE, a computational framework designed for multislice joint embedding, spatial domain identification, and gene expression imputation. Specifically, MUSE integrates a two-module architecture to ensure robust cross-slice alignment and data harmonization. The alignment module models each slice as a graph and employs optimal transport to align cells across slices while preserving spatial continuity. The optimization module further refines integration by incorporating an alignment loss, allowing lower-quality data to leverage structural information from higher-quality slices. Additionally, MUSE generates virtual neighbors from aligned cells, enriching contextual information and mitigating data sparsity. These design principles enable seamless integration with existing single-slice methods, extending their applicability to multi-slice ST analysis. To comprehensively evaluate its performance, we applied MUSE to 12 real and 48 simulated datasets spanning a range of data qualities. Across all metrics, MUSE consistently outperformed existing methods in cross-slice consistency, spatial domain identification, and gene expression imputation. To promote accessibility and adoption, we provide MUSE as an open-source software package. As multi-slice ST datasets become increasingly prevalent, MUSE provides a robust and extensible framework designed to effectively integrate growing numbers of slices, thereby advancing the analysis of tissue architectures and spatial gene expression in complex biological systems.
Temporal Graph Neural Networks (TGNNs) have become indispensable for analyzing dynamic graphs in critical applications such as social networks, communication systems, and financial networks. However, the robustness of TGNNs against adversarial attacks, particularly sophisticated attacks that exploit the temporal dimension, remains a significant challenge. Existing attack methods for Spatio-Temporal Dynamic Graphs (STDGs) often rely on simplistic, easily detectable perturbations (e.g., random edge additions/deletions) and fail to strategically target the most influential nodes and edges for maximum impact. We introduce the High Impact Attack (HIA), a novel restricted black-box attack framework specifically designed to overcome these limitations and expose critical vulnerabilities in TGNNs. HIA leverages a data-driven surrogate model to identify structurally important nodes (central to network connectivity) and dynamically important nodes (critical for the graph's temporal evolution). It then employs a hybrid perturbation strategy, combining strategic edge injection (to create misleading connections) and targeted edge deletion (to disrupt essential pathways), maximizing TGNN performance degradation. Importantly, HIA minimizes the number of perturbations to enhance stealth, making it more challenging to detect. Comprehensive experiments on five real-world datasets and four representative TGNN architectures (TGN, JODIE, DySAT, and TGAT) demonstrate that HIA significantly reduces TGNN accuracy on the link prediction task, achieving up to a 35.55% decrease in Mean Reciprocal Rank (MRR) - a substantial improvement over state-of-the-art baselines. These results highlight fundamental vulnerabilities in current STDG models and underscore the urgent need for robust defenses that account for both structural and temporal dynamics. Code and Data are available at https://github.com/ryandhjeon/hia.
Large pre-trained models have achieved notable success across a range of downstream tasks. However, recent research shows that a type of adversarial attack (i.e., backdoor attack) can manipulate the behavior of machine learning models through contaminating their training dataset, posing significant threat in the real-world application of large pre-trained model, especially for those customized models. Therefore, addressing the unique challenges for exploring vulnerability of pre-trained models is of paramount importance. Through empirical studies on the capability for performing backdoor attack in large pre-trained models (e.g., ViT), we find the following unique challenges of attacking large pre-trained models: 1) the inability to manipulate or even access large training datasets, and 2) the substantial computational resources required for training or fine-tuning these models. To address these challenges, we establish new standards for an effective and feasible backdoor attack in the context of large pre-trained models. In line with these standards, we introduce our EDT model, an Efficient, Data-free, Training-free backdoor attack method. Inspired by model editing techniques, EDT injects an editing-based lightweight codebook into the backdoor of large pre-trained models, which replaces the embedding of the poisoned image with the target image without poisoning the training dataset or training the victim model. Our experiments, conducted across various pre-trained models such as ViT, CLIP, BLIP, and stable diffusion, and on downstream tasks including image classification, image captioning, and image generation, demonstrate the effectiveness of our method. Our code is available at https://github.com/donglgcn/Editing/.
Spatial transcriptomics has transformed genomic research by measuring spatially resolved gene expressions, allowing us to investigate how cells adapt to their microenvironment via modulating their expressed genes. This essential process usually starts from cell-cell communication (CCC) via ligand-receptor (LR) interaction, leading to regulatory changes within the receiver cell. However, few methods were developed to connect them to provide biological insights into intercellular regulation. To fill this gap, we propose iMiracle, an iterative multi-view graph neural network that models each cell's intercellular regulation with three key features. Firstly, iMiracle integrates inter- and intra-cellular networks to jointly estimate cell-type- and micro-environment-driven gene expressions. Optionally, it allows prior knowledge of intra-cellular networks as pre-structured masks to maintain biological relevance. Secondly, iMiracle employs iterative learning to overcome the sparsity of spatial transcriptomic data and gradually fill in the missing edges in the CCC network. Thirdly, iMiracle infers a cell-specific ligand-gene regulatory score based on the contributions of different LR pairs to interpret inter-cellular regulation. We applied iMiracle to nine simulated and eight real datasets from three sequencing platforms and demonstrated that iMiracle consistently outperformed ten methods in gene expression imputation and four methods in regulatory score inference. Lastly, we developed iMiracle as an open-source software and anticipate that it can be a powerful tool in decoding the complexities of inter-cellular transcriptional regulation.
Heterogeneous networks contain multiple types of nodes and links, with some link types encapsulating hierarchical structure over entities. Hierarchical relationships can codify information such as subcategories or one entity being subsumed by another and are often used for organizing conceptual knowledge into a tree-structured graph. Hyperbolic embedding models learn node representations in a hyperbolic space suitable for preserving the hierarchical structure. Unfortunately, current hyperbolic embedding models only implicitly capture the hierarchical structure, failing to distinguish between node types, and they only assume a single tree. In practice, many networks contain a mixture of hierarchical and non-hierarchical structures, and the hierarchical relations may be represented as multiple trees with complex structures, such as sharing certain entities. In this work, we propose a new hyperbolic representation learning model that can handle complex hierarchical structures and also learn the representation of both hierarchical and non-hierarchic structures. We evaluate our model on several datasets, including identifying relevant articles for a systematic review, which is an essential tool for evidence-driven medicine and node classification.
Federated learning on graphs (a.k.a., federated graph learning - FGL) has recently received increasing attention due to its capacity to enable collaborative learning over distributed graph datasets without compromising local clients' data privacy. In previous works, clients of FGL typically represent institutes or organizations that possess sets of entire graphs (e.g., molecule graphs in biochemical research) or parts of a larger graph (e.g., sub-user networks of e-commerce platforms). However, another natural paradigm exists where clients act as remote devices retaining the graph structures of local neighborhoods centered around the device owners (i.e., ego-networks), which can be modeled for specific graph applications such as user profiling on social ego-networks and infection prediction on contact ego-networks. FGL in such novel yet realistic ego-network settings faces the unique challenge of incomplete neighborhood information for non-ego local nodes since they likely appear and have different sets of neighbors in multiple ego-networks. To address this challenge, we propose an FGL method for distributed ego-networks in which clients obtain complete neighborhood information of local nodes through sharing node embeddings with other clients. A contrastive learning mechanism is proposed to bridge the gap between local and global node embeddings and stabilize the local training of graph neural network models, while a secure embedding sharing protocol is employed to protect individual node identity and embedding privacy against the server and other clients. Comprehensive experiments on various distributed ego-network datasets successfully demonstrate the effectiveness of our proposed embedding sharing method on top of different federated model sharing frameworks, and we also provide discussions on the potential efficiency and privacy drawbacks of the method as well as their future mitigation.
Single-cell sequencing technologies have revolutionized genomics by enabling the simultaneous profiling of various molecular modalities within individual cells. Their integration, especially cross-modality translation, offers deep insights into cellular regulatory mechanisms. Many methods have been developed for cross-modality translation, but their reliance on scarce high-quality co-assay data limits their applicability. Addressing this, we introduce scACT, a deep generative model designed to extract cross-modality biological insights from unpaired single-cell data. scACT tackles three major challenges: aligning unpaired multi-modal data via adversarial training, facilitating cross-modality translation without prior knowledge via cycle-consistent training, and enabling interpretable regulatory interconnections explorations via in-silico perturbations. To test its performance, we applied scACT on diverse single-cell datasets and found it outperformed existing methods in all three tasks. Finally, we have developed scACT as an individual open-source software package to advance single-cell omics data processing and analysis within the research community.
The shuffle model of Differential Privacy (DP) is an enhanced privacy protocol which significantly amplifies the central DP guarantee by anonymizing and shuffling the local randomized data. Yet, deriving a tight privacy bound is challenging due to its complicated randomization protocol. While most existing works focused on uniform local privacy settings, this work focuses on a more practical personalized privacy setting. To bound the privacy after shuffling, we need to capture the probability of each user generating clones of the neighboring data points and quantify the indistinguishability between two distributions of the number of clones on neighboring datasets. Existing works either inaccurately capture the probability or underestimate the indistinguishability. We develop a more precise analysis, which yields a general and tighter bound for arbitrary DP mechanisms. Firstly, we derive the clone-generating probability by hypothesis testing, which leads to a more accurate characterization of the probability. Secondly, we analyze the indistinguishability in the context of f -DP, where the convexity of the distributions is leveraged to achieve a tighter privacy bound. Theoretical and numerical results demonstrate that our bound remarkably outperforms the existing results in the literature. The code is publicly available at https://github.com/Emory-AIMS/HPS.git.
Spatiotemporal time series are usually collected via monitoring sensors placed at different locations, which usually contain missing values due to various failures, such as mechanical damages and Internet outages. Imputing the missing values is crucial for analyzing time series. When recovering a specific data point, most existing methods consider all the information relevant to that point regardless of the cause-and-effect relationship. During data collection, it is inevitable that some unknown confounders are included, e.g., background noise in time series and non-causal shortcut edges in the constructed sensor network. These confounders could open backdoor paths and establish non-causal correlations between the input and output. Over-exploiting these non-causal correlations could cause overfitting. In this paper, we first revisit spatiotemporal time series imputation from a causal perspective and show how to block the confounders via the frontdoor adjustment. Based on the results of frontdoor adjustment, we introduce a novel Causality-Aware Spatiotemporal Graph Neural Network (Casper), which contains a novel Prompt Based Decoder (PBD) and a Spatiotemporal Causal Attention (SCA). PBD could reduce the impact of confounders and SCA could discover the sparse causal relationships among embeddings. Theoretical analysis reveals that SCA discovers causal relationships based on the values of gradients. We evaluate Casper on three real-world datasets, and the experimental results show that Casper could outperform the baselines and could effectively discover the causal relationships.
The exponential growth in scholarly publications necessitates advanced tools for efficient article retrieval, especially in interdisciplinary fields where diverse terminologies are used to describe similar research. Traditional keyword-based search engines often fall short in assisting users who may not be familiar with specific terminologies. To address this, we present a knowledge graph based paper search engine for biomedical research to enhance the user experience in discovering relevant queries and articles. The system, dubbed DiscoverPath, employs Named Entity Recognition (NER) and part-of-speech (POS) tagging to extract terminologies and relationships from article abstracts to create a KG. To reduce information overload, DiscoverPath presents users with a focused subgraph containing the queried entity and its neighboring nodes and incorporates a query recommendation system enabling users to iteratively refine their queries. The system is equipped with an accessible Graphical User Interface that provides an intuitive visualization of the KG, query recommendations, and detailed article information, enabling efficient article retrieval, thus fostering interdisciplinary knowledge exploration.
Predicting adverse drug reactions (ADRs) of drugs is one of the most critical steps in drug development. By pre-estimating the adverse reactions, researchers and drug development companies can greatly prevent the potential ADR risks and tragedies. However, the current ADR prediction methods suffer from several limitations. First, the prediction results are based on pure drug-related information, which makes them impossible to be directly applied for the personalized ADR prediction task. The lack of personalization of models also makes rare adverse events hard to be predicted. Therefore, it is of great interest to develop a new personalized ADR prediction method by introducing additional sources, e.g., patient health records. However, few methods have tried to use additional sources. In the meantime, the variety of different source formats and structures makes this task more challenging. To address the above challenges, we propose a novel personalized multi-sourced-based drug adverse reaction prediction model named pADR. pADR first works on every single source to transform them into proper representations. Next, a hierarchical multi-sourced Transformer is designed to automatically model the interactions between different sources and fuse them together for the final adverse event prediction. Experimental results on a new multi-sourced ADR prediction dataset show that pADR1 outperforms state-of-the-art drug-based baselines. Moreover, the case and ablation studies also illustrate the effectiveness of our proposed fusion strategies and the reasonableness of each module design.
Clinical trials aim to study new tests and evaluate their effects on human health outcomes, which has a huge market size. However, carrying out clinical trials is expensive and time-consuming and often ends in no results. It will revolutionize clinical practice if we can develop an effective model to automatically estimate the status of a clinical trial and find out possible failure reasons. However, it is challenging to develop such a model because of the lack of a benchmark dataset. To address these challenges, in this paper, we first build a new dataset by extracting the publicly available clinical trial reports from ClinicalTrials.gov. The associated status of each report is treated as the status label. To analyze the failure reasons, domain experts help us manually annotate each failed report based on the description associated with it. More importantly, we examine several state-of-the-art text classification baselines on this task and find out that the unique format of the clinical trial protocols plays an essential role in affecting prediction accuracy, demonstrating the need for specially designed clinical trial classification models.
Continual learning has emerged as a powerful approach to address the challenges of non-stationary environments, allowing machine learning models to adapt to new data while retaining the previously acquired knowledge. In time-sensitive healthcare applications, where entities such as physicians, hospital rooms, and medications exhibit continuous changes over time, continual learning holds great promise, yet its application remains relatively unexplored. This paper aims to bridge this gap by proposing a novel framework, i.e., Continually-Adaptive Representation Learning, designed to adapt representations in response to changing data distributions in evolving healthcare applications. Specifically, the proposed approach develops a continual learning strategy wherein the context information (e.g., interactions) of healthcare entities is exploited to continually identify and retrain the representations of those entities whose context evolved over time. Moreover, different from existing approaches, the proposed approach leverages the valuable patient information present in clinical notes to generate accurate and robust healthcare embeddings. Notably, the proposed continually-adaptive representations have practical benefits in low-resource clinical settings where it is difficult to training machine learning models from scratch to accommodate the newly available data streams. Experimental evaluations on real-world healthcare datasets demonstrate the effectiveness of our approach in time-sensitive healthcare applications such as Clostridioides difficile (C.diff) Infection (CDI) incidence prediction task and medical intensive care unit transfer prediction task.
Sharing health data is vital in advancing medical research and transforming knowledge into clinical practice. Meanwhile, protecting the privacy of data contributors is of paramount importance. To that end, several privacy approaches have been proposed to protect individual data contributors in data sharing, including data anonymization and data synthesis techniques. These approaches have shown promising results in providing privacy protection at the dataset level. In this work, we study the privacy challenges in enabling fine-grained privacy in health data sharing. Our work is motivated by recent research findings, in which patients and healthcare providers may have different privacy preferences and policies that need to be addressed. Specifically, we propose a novel and effective privacy solution that enables data curators (e.g., healthcare providers) to protect sensitive data elements while preserving data usefulness. Our solution builds on randomized techniques to provide rigorous privacy protection for sensitive elements and leverages graphical models to mitigate privacy leakage due to dependent elements. To enhance the usefulness of the shared data, our randomized mechanism incorporates domain knowledge to preserve semantic similarity and adopts a block-structured design to minimize utility loss. Evaluations with real-world health data demonstrate the effectiveness of our approach and the usefulness of the shared data for health applications.
Healthcare providers generate a medical claim after every patient visit. A medical claim consists of a list of medical codes describing the diagnosis and any treatment provided during the visit. Medical claims have been popular in medical research as a data source for retrospective cohort studies. This paper introduces a medical claim visualization system (MedCV) that supports cohort selection from medical claim data. MedCV was developed as part of a design study in collaboration with clinical researchers and statisticians. It helps a researcher to define inclusion rules for cohort selection by revealing relationships between medical codes and visualizing medical claims and patient timelines. Evaluation of our system through a user study indicates that MedCV enables domain experts to define high-quality inclusion rules in a time-efficient manner.
With the ever-increasing abundance of biomedical articles, improving the accuracy of keyword search results becomes crucial for ensuring reproducible research. However, keyword extraction for biomedical articles is hard due to the existence of obscure keywords and the lack of a comprehensive benchmark. PubMedAKE is an author-assigned keyword extraction dataset that contains the title, abstract, and keywords of over 843,269 articles from the PubMed open access subset database. This dataset, publicly available on Zenodo, is the largest keyword extraction benchmark with sufficient samples to train neural networks. Experimental results using state-of-the-art baseline methods illustrate the need for developing automatic keyword extraction methods for biomedical literature.
Representation learning on static graph-structured data has shown a significant impact on many real-world applications. However, less attention has been paid to the evolving nature of temporal networks, in which the edges are often changing over time. The embeddings of such temporal networks should encode both graph-structured information and the temporally evolving pattern. Existing approaches in learning temporally evolving network representations fail to capture the temporal interdependence. In this paper, we propose Toffee, a novel approach for temporal network representation learning based on tensor decomposition. Our method exploits the tensor-tensor product operator to encode the cross-time information, so that the periodic changes in the evolving networks can be captured. Experimental results demonstrate that Toffee outperforms existing methods on multiple real-world temporal networks in generating effective embeddings for the link prediction tasks.
Video accessibility is crucial for blind screen-reader users as online videos are increasingly playing an essential role in education, employment, and entertainment. While there exist quite a few techniques and guidelines that focus on creating accessible videos, there is a dearth of research that attempts to characterize the accessibility of existing videos. Therefore in this paper, we define and investigate a diverse set of video and audio-based accessibility features in an effort to characterize accessible and inaccessible videos. As a ground truth for our investigation, we built a custom dataset of 600 videos, in which each video was assigned an accessibility score based on the number of its wins in a Swiss-system tournament, where human annotators performed pairwise accessibility comparisons of videos. In contrast to existing accessibility research where the assessments are typically done by blind users, we recruited sighted users for our effort, since videos comprise a special case where sight could be required to better judge if any particular scene in a video is presently accessible or not. Subsequently, by examining the extent of association between the accessibility features and the accessibility scores, we could determine the features that significantly (positively or negatively) impact video accessibility and therefore serve as good indicators for assessing the accessibility of videos. Using the custom dataset, we also trained machine learning models that leveraged our handcrafted features to either classify an arbitrary video as accessible/inaccessible or predict an accessibility score for the video. Evaluation of our models yielded an F-1 score of 0.675 for binary classification and a mean absolute error of 0.53 for score prediction, thereby demonstrating their potential in video accessibility assessment while also illuminating their current limitations and the need for further research in this area.
Learning of classification models from real-world data often requires substantial human effort devoted to instance annotation. As this process can be very time-consuming and costly, finding effective ways to reduce the annotation cost becomes critical for building such models. To address this problem we explore a new type of human feedback - region-based feedback. Briefly, a region is defined as a hypercubic subspace of the input data space and represents a subpopulation of data instances; the region's label is a human assessment of the class proportion of the data subpopulation. By using learning from label proportions algorithms one can learn instance-based classifiers from such labeled regions. In general, the key challenge is that there can be infinite many regions one can define and query in a given data space. To minimize the number and complexity of region-based queries, we propose and develop a hierarchical active learning solution that aims at incrementally building a concise hierarchy of regions. Furthermore, to avoid building a possibly class-irrelevant region hierarchy, we further propose to grow multiple different hierarchies in parallel and expand those more informative hierarchies. Through experiments on numerous data sets, we demonstrate that methods using region-based feedback can learn very good classifiers from very few and simple queries, and hence are highly effective in reducing human annotation effort needed for building classification models.
Tensor factorization has been demonstrated as an efficient approach for computational phenotyping, where massive electronic health records (EHRs) are converted to concise and meaningful clinical concepts. While distributing the tensor factorization tasks to local sites can avoid direct data sharing, it still requires the exchange of intermediary results which could reveal sensitive patient information. Therefore, the challenge is how to jointly decompose the tensor under rigorous and principled privacy constraints, while still support the model's interpretability. We propose DPFact, a privacy-preserving collaborative tensor factorization method for computational phenotyping using EHR. It embeds advanced privacy-preserving mechanisms with collaborative learning. Hospitals can keep their EHR database private but also collaboratively learn meaningful clinical concepts by sharing differentially private intermediary results. Moreover, DPFact solves the heterogeneous patient population using a structured sparsity term. In our framework, each hospital decomposes its local tensors and sends the updated intermediary results with output perturbation every several iterations to a semi-trusted server which generates the phenotypes. The evaluation on both real-world and synthetic datasets demonstrated that under strict privacy constraints, our method is more accurate and communication-efficient than state-of-the-art baseline methods.
