Session Abstracts
- Dr Monica Agrawal – Clinical Foundation Models
- Dr Heng Huang – Advanced Machine Learning Enabled Imaging-Omics Analytics
- Dr Bang Liu – Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems
- Dr Jie Liu – Foundational Models and Knowledge Graphs for Consolidating our Knowledge Regarding the Human Genome
- Dr Sriram Sankararaman – Machine Learning for Large-Scale, Multimodal, Biomedical Data
- Dr Fei Wang – Harnessing Real World Evidence for Better Clinical Trials with AI
- Dr May Dongmei Wang – Challenges and Opportunities in Translating AI for Healthcare
- Dr Wei Wang – Medical Information Retrieval in the Era of Large Language Models (LLMs)
- Dr Qiao Liu, Dr Huaxiu Yao, Dr. Xin Wang, Dr Hongtu Zhu – Causal Generalist Medical AI
Clinical Foundation Models
Dr Monica Agrawal
Foundation models, particularly large language models, are changing how we process, interpret, and predict from clinical data. In this short course, you will learn the paradigms and training behind clinical foundation models, including LLMs, visual language models, and others. We will cover various techniques including fine-tuning and weak supervision, as well describe the evaluation landscape.
Advanced Machine Learning Enabled Imaging-Omics Analytics
Dr Heng Huang
Machine learning is accelerating the translation of biological and biomedical data to advance the detection, diagnosis, treatment, and prevention of diseases. However, the unprecedented scale and complexity of large-scale biomedical data have presented critical computational bottlenecks requiring new concepts and enabling tools. To address the challenging problems in the emerging biomedical and health applications, the novel nonlinear machine learning models have been designed with theoretical foundations for scalable nonlinear association study, interactive feature detection, multi-dimensional data integration, longitudinal feature learning, etc. I will show that how to design these large-scale nonlinear machine learning models and apply them to analyze the multi-modal and longitudinal neuroimaging and genome-wide array data in imaging genomics and discover the phenotypic and genotypic biomarkers to characterize the neurodegenerative process in the progression of Alzheimer’s disease, analyze the retinal imaging omics data to detect the biomarkers for ocular diseases, identify the histopathological image markers and the multi-dimensional cancer genomic biomarkers in The Cancer Genome Atlas (TCGA) for precision medicine, and analyze the single-cell multi-omics to understand the biological mechanisms.
Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems
Dr. Bang Liu
The advent of large language models (LLMs) has catalyzed a transformative shift in artificial intelligence, paving the way for advanced intelligent agents capable of sophisticated reasoning, robust perception, and versatile action across diverse domains. As these agents increasingly drive AI research and practical applications, their design, evaluation, and continuous improvement present intricate, multifaceted challenges. This short course aims to provide a comprehensive overview, framing intelligent agents within a modular, brain-inspired architecture that integrates principles from cognitive science, neuroscience, and computational research.
We structure our course into four interconnected parts. First, we delve into the modular foundations of intelligent agents, systematically mapping their cognitive, perceptual, and operational modules onto analogous human brain functionalities, and elucidating core components such as memory, world modeling, reward processing, and emotion-like systems. Second, we discuss self-enhancement and adaptive evolution mechanisms, exploring how agents autonomously refine their capabilities, adapt to dynamic environments, and achieve continual learning through automated optimization paradigms, including emerging AutoML and LLM-driven optimization strategies. Third, we examine collaborative and evolutionary multi-agent systems, investigating the collective intelligence emerging from agent interactions, cooperation, and societal structures, highlighting parallels to human social dynamics. Finally, we address the critical imperative of building safe, secure, and beneficial AI systems, emphasizing intrinsic and extrinsic security threats, ethical alignment, robustness, and practical mitigation strategies necessary for trustworthy real-world deployment.
By synthesizing modular AI architectures with insights from neural and cognitive sciences, this short course aims to identify key research gaps, challenges, and opportunities, offering a unified, interdisciplinary roadmap, guiding the development of more powerful, adaptive, and socially aligned intelligent agents with meaningful societal benefit.
Foundational Models and Knowledge Graphs for Consolidating our Knowledge Regarding the Human Genome
Dr. Jie Liu
Our knowledge regarding the human genome has been exponentially increasing. The knowledge presents in different formats, including direct measurements of genomic entities with the ever-evolving biotechnologies, annotations by groups of experts from different consortia, discoveries from individual studies published as free text in biomedical literature, and insights learned from computational models trained on large-scale genomic datasets. However, we currently do not have an infrastructure to consolidate these heterogeneous knowledge sources. As a result, genomic researchers nowadays spend increasingly more time searching for relevant datasets and literature for scientific discoveries, annotations and conclusions, and unfortunately they do not have AI-powered tools to navigate existing knowledge and prioritize their hypotheses and research activities. In this talk, I will describe deep learning models and knowledge graphs from my lab for consolidating our knowledge regarding the human genome. The deep learning models include CAESAR and EPCOT. The knowledge graphs include GenomicKB and Genomic Literature Knowledge Base. Our works not only have an enormous positive impact on sharing genomic knowledge and facilitating new genomic knowledge discovery, they would also help to promote open science, inclusivity and fairness in the areas of computational genomics and data science.
Machine Learning for Large-Scale, Multimodal, Biomedical Data
Dr Sriram Sankararaman
The quest to understand biological underpinnings of disease has been revolutionized by the collection of rich datasets that span multiple modalities (phenotypes, genetics, structured EHR, imaging, time-series data) across millions of individuals in diverse populations in repositories that are termed Biobanks. However analyses of these Biobank-scale datasets present substantial statistical and computational challenges.
I will describe new machine learning models that enable design new analyses at scale for a suite of problems that arise in the analysis of Biobank data: highly scalable machine learning models that enable the detection of complex associations between genes and diseases, deep-learning based phenotype imputation to deal with complex patterns of missingness, and scalable train-free approaches to deal with 3D medical volumes. I will show how these models can be applied to about half a million individuals from the UK Biobank to obtain novel insights into genetic associations for traits, to identify new genes that confer risk for hard-to-measure diseases, and to learn more meaningful representations of brain MRI volumes.
Harnessing Real World Evidence for Better Clinical Trials with AI
Dr Fei Wang
Clinical trial is a time consuming and expensive process in drug development. It is estimated that on average it takes over 1 billion dollars and 10-15 years for a successful drug to go through the different phases of clinical trials until receiving FDA approval. The availability of large-scale real-world data, especially electronic health records and pharmaceutical/insurance claims, along with the advancement of AI technologies, offer an unprecedented opportunity for making clinical trials more effective and efficient with the real world evidence. This short course will introduce what are clinical trials and how real world evidence can be harnessed from the real world data with AI, as well as the potential challenges and future directions associated with this process.
Challenges and Opportunities in Translating AI for Healthcare
Dr. May Dongmei Wang
The 21st century has witnessed major challenges caused by both COVID19 pandemic and aging society. I will discuss the grand challenges and opportunities in AI for healthcare. In addition, I will present AI Foundation Models, AI Implementation Science, and Metaverse for Healthcare. OpenAI LLMs are susceptible to hallucinated information and lack of logical reasoning. Thus, the healthcare LLMs require more domain expertise, patient specific data, logical reasoning to handle complex inferences, and computation and transparency for broad adoption in clinical settings. Working with Microsoft Accelerating Foundation Models Research, we developed the first retrieval augmented generation (RAG) solutions that augment LLMs with the most recent domain-specific medical knowledge for clinics, followed by EHRAgent published in Association for Computational Linguistics and Empirical Methods in Natural Language Processing. We formulated a clinical problem-solving process as an executable action sequence code plan with a code executor to get environment feedback to improve code generation for tabular reasoning tasks. Our AI Implementation Science projects were accepted into AMIA 3-Tier AI Showcase, and our real-time Digital Twin and Metaverse for rehabilitation were published in IEEE International Conference in Intelligent Reality.
Medical Information Retrieval in the Era of Large Language Models (LLMs)
Dr. Wei Wang
The emergence of large language models (LLMs) has introduced a new paradigm in data modeling. These models replace specialized models designed for individual tasks with unified models that are effective across a broad range of problems. In biomedical domains, this shift not only transforms approaches to handling natural language tasks (e.g., scientific papers) but also suggests new methods for dealing with other data types (e.g., molecules, proteins, pathology images). In many fields, LLM has already shown great potential to accelerate scientific discovery. In this talk, I will present our recent work on LLMs, especially in the context of science and engineering research.
Causal Generalist Medical AI
Dr Qiao Liu, Dr Huaxiu Yao, Dr Xin Wang, Dr Hongtu Zhu
The rapid evolution of flexible and reusable artificial intelligence (AI) models is transforming medical science. This short course introduces Causal Generalist Medical AI (Causal GMAI)—a paradigm that integrates causal inference with generalist AI models to enhance interpretability, robustness, and generalizability in medical decision-making. Causal GMAI employs self-supervised, semi-supervised, and supervised learning on diverse multimodal datasets—including imaging, electronic health records, clinical trials, laboratory results, genomics, knowledge graphs, and medical text—to perform a wide range of tasks with minimal task-specific supervision. By embedding causal reasoning, these models go beyond prediction to infer underlying causal relationships, improving diagnostic accuracy, treatment recommendations, and personalized medicine. The course covers key technical components such as causal discovery, counterfactual reasoning, and domain adaptation, alongside real-world applications. We will also explore challenges in regulation, validation, and dataset curation to ensure clinical reliability and ethical deployment. Designed for researchers, clinicians, data scientists, and AI practitioners, this course provides a foundation for advancing the next generation of trustworthy and interpretable medical AI.
