This paper significantly improves how scientists analyze complex molecular movements, like protein folding, using a technique that simplifies these motions into transitions between 'states.' It shows that by using two specialized mathematical descriptions instead of one.
Markov state models revisited: Principles and algorithms for unbiased observables · David Aristof, Robert J. Webber, Daniel M. Zuckerman · arXiv:2607.19452
AI models are great at understanding DNA, but we often don't know how they reach their conclusions. A way to peer inside these models, identifying specific internal "thoughts" that correspond to known biological signals, like the DNA sequences that turn genes on or off.
Causal dictionary learning reveals and validates transcription-factor binding features in genomic language models · Sarwan Ali · arXiv:2607.19618
A new software method that can turn regular, fast 2D MRI scans into detailed 4D "movies" showing how organs move, like with breathing.
Portable surrogate-free 4D MRI from standard fast multi-slice 2D MRI via implicit neural representations · Muheng Li, Xinyang Wu, Xia Li et al. · arXiv:2607.17906
A new method to automatically create highly accurate computer models of individual hearts by directly fitting them to real-time electrical activity measurements. It uses advanced AI-like techniques to adjust the model's complex equations until its simulated electrical signals precisely match a patient's heart rhythm, even from noisy data.
Differentiable Cardiac Electrophysiology Simulations for Dynamical State and Parameter Estimation · Adarsh Pashikanti, Shrey Chowdhary, Alex Ho et al. · arXiv:2607.15492
This paper uses machine learning to predict how strongly specific proteins will stick to different parts of our DNA, essentially creating a "binding map" across the entire genome. They then use this tool to precisely map where a key gene-regulating protein, PU.1, prefers to bind.
The energy landscape of DNA-binding proteins along the genome · Alessandro Pandolfi, Riccardo Beccaria, Guido Tiana · arXiv:2607.17753