Computational science articles from across Nature Portfolio

Understanding how people move through cities is essential for public health and urban planning, yet most cities worldwide lack reliable mobility data. A new model reconstructs these movement patterns from openly available maps and population data, revealing that income inequality shapes urban mobility in ways that transcend local geography.

Free boundary problems, such as modelling glacier melt, are difficult to capture with neural operators. A new framework addresses this challenge by leveraging the mathematical principle of topological conjugacy.

The BRIDGE benchmark systematically evaluated the performance of 95 large language models (LLMs) on real-world clinical data on tasks spanning the patient care continuum. This benchmark, which is being continuously updated, provides a field reference for improving how LLMs understand clinical texts.

A large language model artificial intelligence agent operating in a sandboxed electronic health record system can autonomously take patient histories, order tests, interpret findings, diagnose conditions and propose treatments, outperforming experienced clinicians while adhering to safety standards and clinical guidelines.

Jhelum Chakravorty and colleagues introduce an end-to-end approach to robust discovery of electrical dynamical systems with a class of polynomial dynamics. This research lays the ground for modelling diverse multi-physics systems with a balance in computation cost and performance.

The authors developed an end to-end AI framework for discovering multiscale mathematical formulas in complex systems, using representation learning and symbolic regression to extract scale-specific laws, improving efficiency and enabling automated scientific discovery beyond single-scale methods.

Pengfei Sun et al. develop a spiking neural network with a dual memory pathway, co-designed with a custom neuromorphic chip. The approach delivers over 4× throughput and 5x energy efficiency gains while using 40–60% fewer parameters than state-of-the-art implementations.

The lattice architecture of the HIV capsid holds the key to understanding its biophysical properties and function. Li et al. introduce a geometric criterion revealing its implications for molecular frustration, lattice anomalies and cofactor binding.

Understanding how people move through cities is essential for public health and urban planning, yet most cities worldwide lack reliable mobility data. A new model reconstructs these movement patterns from openly available maps and population data, revealing that income inequality shapes urban mobility in ways that transcend local geography.

Nature Computational Science now takes part in the co-review initiative in order to encourage the involvement of early-career researchers in the peer review process and to ensure proper recognition of their service.

Free boundary problems, such as modelling glacier melt, are difficult to capture with neural operators. A new framework addresses this challenge by leveraging the mathematical principle of topological conjugacy.

Explaining the global decision logic of black-box medical artificial intelligence (AI) models is a formidable challenge. We proposed class-association manifold learning, an explanatory framework that harnesses low-dimensional manifolds to visualize and accurately explore hidden global decision rules captured by medical AI models. Class-association manifold learning enabled human-interpretable medical knowledge discovery while ensuring AI model alignment.