Minisymposium

Point mutations in the protein sequences are known to have potential to alter the protein’s native fold, stability, and functions, and may result in observable disease phenotypes. Currently, there are over 200,000 experimental protein structures deposited in the Protein Data Bank, enabling the training of deep learning models for structure prediction. However, these models often fail to accurately predict the structural effects of mutations. As a result, we lack a quantitative understanding of the effect of mutations on protein structures. Here, we curate a dataset of x-ray crystal structure duplicates and their corresponding single-point mutant structures, creating opportunities to leverage high-performance computing for large-scale analysis and the training of predictive models. We quantify the local structural deformation per residue between wildtype-mutant pairs and compare them to the baseline within wildtype duplicates. Our analysis shows that on average, the magnitude of structural perturbation decreases as the sequence and spatial distance from the mutation site increase. We aim to illustrate the key physical features that determine the mutational impact and develop predictive models using data-driven approach in future studies. These results could advance our understanding of genetic diseases and support the development of structure-based drug discovery and therapeutic design.

Despite the high-dimensionality of images, the sets of images of 3D objects have long been hypothesized to form low-dimensional manifolds. What is the nature of such manifolds? How do they differ across objects and object classes? Answering these questions can provide key insights in explaining and advancing success of machine learning algorithms in computer vision. This paper investigates dual tasks -- learning and analyzing shapes of image manifolds -- by revisiting a classical problem of manifold learning but from a novel geometrical perspective. It uses geometry-preserving transformations to map the pose image manifolds, sets of images formed by rotating 3D objects, to low-dimensional latent spaces. The pose manifolds of different objects in latent spaces are found to be nonlinear, smooth manifolds. The paper then compares shapes of these manifolds for different objects using Kendall's shape analysis, modulo rigid motions and global scaling, and clusters objects according to these shape metrics. Interestingly, pose manifolds for objects from the same classes are frequently clustered together. The geometries of image manifolds can be exploited to simplify vision and image processing tasks, to predict performances, and to provide insights into learning methods.

African Swine Fever (ASF) is a highly contagious and deadly viral disease infecting domestic and feral swine populations in Africa and Asia and more recently in the Europe, South America, and Caribbean. ASF has devastating impacts on swine industries in the affected countries. This study proposes to develop a ASF farmer vulnerability risk index for rural swine farms that integrates an array of potential environmental factors (e.g., domestic and feral swine population densities, precipitation, temperature, vegetation), geographic and social factors and seasonality to assess and predict regions where outbreaks are more likely to occur.   The approach is data agnostic, leveraging a broad range of available data with the goal of identifying discriminating features in a data-driven manner.   A feature indexing approach is used to construct labeled training data from historical outbreaks to train a machine learning model to produce a spatial risk index.   Sparse optimization tools are employed to identify the salient features most useful for predictive modeling. The resultant risk index can guide surveillance and preventive strategies, while also outlining limitations related to data granularity and model generalizability. This study incorporates integrating diverse data set to identify areas of risk to potentially inform mitigating the spread of ASF.

Understanding the genetic and molecular underpinnings of addiction and related disorders requires integrative approaches that leverage large-scale omics data, network biology, and artificial intelligence. This work presents a systems biology framework that combines predictive expression networks, foundation models, and AI agents to elucidate mechanisms underlying opioid and nicotine addiction. By integrating genome-wide association studies (GWAS), transcriptomics, and multiplex network modeling, we identify gene clusters linked to addiction-related phenotypes, emphasizing shared mechanisms between opioid use and smoking cessation. Using the MENTOR framework, we partition genes of interest into mechanistically coherent clades, revealing significant overlap between addiction pathways. Network-based analyses uncover key regulators, including BDNF/NTRK2 and MAPK signaling, which influence neuronal plasticity and reinforcement learning. AI-driven interpretation automates gene-function annotation, improving mechanistic inference. Further, retrieval-augmented generation (RAG) agents and reinforcement learning models facilitate high-throughput interpretation of biological networks, accelerating hypothesis generation. This study highlights AI’s role in translating multi-omics data into actionable insights for addiction biology. The framework extends to broader disease contexts, offering a scalable model for systems medicine. Future directions include validation through retrospective clinical trials and experimental assays, emphasizing the potential for AI-guided therapeutic discovery.