A standardized way to describe complex scientific models that combine established equations (like those describing how things change over time) with modern machine learning.
PEtab SciML: an exchange format for specifying and training dynamic scientific machine learning models · Sebastian Persson, Branwen Snelling, Maren Philipps et al. · arXiv:2608.20184
This paper shows how a smart machine learning approach, called Bayesian neural networks, can effectively design proteins to have several desired traits at once.
Multitask Bayesian Neural Networks for Multiparameter Protein Engineering · Fabio Herrera-Rocha, David Medina-Ortiz, Desiree Wyrzykala et al. · arXiv:2608.18604
A new way to determine the 3D atomic structure of proteins from cryo-electron microscopy images. Instead of first building a blurry 3D image of the molecule, they directly reconstruct a protein's 'skeleton' by mathematically adjusting a guessed shape until its simulated 2D views match the experimental data.
Recovering protein conformations from single-particle cryo-EM data via indirect shape matching gradient flows · Erik Jansson, Jonathan Krook, Ozan Öktem et al. · arXiv:2608.17759
GENIE, an AI model that learns from detailed simulations to quickly and accurately forecast epidemic trends, like hospitalizations, at a very local level.
GENIE: Generative Neural Inference for Epidemics · Laura M. Guzmán-Rincón, George R. E. Bradley, Joel Kandiah et al. · arXiv:2608.20253
A new AI method that more effectively outlines thin, curvy structures like blood vessels in 3D medical scans. It works by smoothly transforming a simple starting shape into the complex vessel network, allowing for fast and accurate identification across different body parts and scan types.
Flow Matching Meets 3D Curvilinear Structure Segmentation in Medical Imaging · Sidi Mohamed Sid'El Moctar, Nicolas Vitry, Hélène Bouvrais · arXiv:2608.19965
A new method that uses detailed brain scans taken over time to generate a continuous score, showing precisely where an individual stands on the Alzheimer's disease spectrum.
Longitudinal Bayesian Learning of Continuous Disease Position across the Alzheimer's Disease Continuum · Yingying Zhang, Kun Zhao, Guodong Liu et al. · arXiv:2608.19436
This paper uses computer models, informed by real patient data, to explore how a new cancer treatment called "adaptive therapy" might perform across different individuals. While adaptive therapy can delay disease progression, it might also increase other risks, like metastasis, if not carefully tailored to each patient's situation.
Adaptive therapy under parametric, structural, and measurement uncertainty · Alexander P Browning, Rebecca M Crossley, Ryan J Murphy et al. · arXiv:2608.18387
An AI model that can accurately reconstruct the complete 3D shape and function of the heart's main pumping chamber from just a few MRI images, rather than the many traditionally needed.
LV-CARE-Diff: A Conditional Anatomy-Aware Diffusion Model for Left Ventricular Shape Reconstruction and Function Quantification from Ultra-Sparse Cine Slices · Xinwang Li, Yu Lian, Bowei Liu et al. · arXiv:2608.17357