Identifying transition state structures in a chemical reaction is important for improving the understanding of the overall reaction mechanism. However, existing methods for transition state structure identification are computationally expensive and tend to have low success rates due to the complexity of potential energy surfaces. In this issue, Chenru Duan et al. introduce a diffusion model that generates chemical reactions in 3D while preserving the desired symmetries. The approach is shown to reduce the transition state search time substantially, from days to seconds. The cover image depicts a potential energy surface for a diffusion process, with transition state structures found on the peaks of the surface.

See Chenru Duan et al.

Granular materials — which correspond to aggregates of closely packed solid particles, such as sand grains — are ubiquitous in our daily lives and important for various applications. In this issue, Lindsay Riley et al. introduce an analysis software that can accurately identify three-dimensional pores in granular systems, that is, the pockets of empty space between packed particles. The tool is able to quantify the solid and void phases from tomography images, shedding light on unknown relationships between particle and pore properties.

See Lindsay Riley et al. and Matthew Evans

Designing molecules with targeted structures and functions is a complex problem that requires expert insights at the molecular level. In this issue, Renana Poranne et al. propose a guided diffusion method for inverse molecular design that combines an equivariant graph neural network and a generative diffusion model. This approach, called GaUDI — a nod to the famous architect Antoni Gaudí — enables the design of polycyclic aromatic systems (depicted on the cover) with improved functionality, a promising result for enhancing the construction and performance of organic electronics.

See Renana Poranne et al. and Ganna Gryn’ova

Spatial urban planning is a highly complex task that mostly depends on the intuition and experiences of human experts. In this issue, Yu Zheng et al. propose an integrated human–ML (machine learning) collaborative workflow that uses deep reinforcement learning to optimize the generation of land use and road layouts for urban communities. Such an approach delegates the more time-consuming and computational-heavy steps to ML, leaving the more conceptual steps in the hands of human planners.

See Yu Zheng et al. and Paolo Santi

In this issue, Matthew D. Witman and colleagues use a graph neural network approach to automate the prediction of defect formation enthalpies in any crystallographic site from an ideal crystal structure. The method has the potential to aid in future materials discovery problems, particularly for materials with clean energy and high-temperature applications, such as the golden crystal structure depicted on the cover, which has a light source in the back representing a high-temperature environment.

See Matthew D. Witman et al.

The field of human mobility has evolved drastically in the past 20 years. The increasing availability of data describing how people move across space and the ever-growing advances in computational science have allowed researchers to uncover regularities in many human activities that involve movements. But what’s next? Laura Alessandretti et al. discuss three key areas in human mobility — framed as minds, societies, and algorithms — where they expect to see substantial improvements in the future. Also in this issue, Marta C. González et al. demonstrate how a combination of both individuals’ mobility data (for instance, from smartphones) and data collected from dwellers (for instance, travel survey data) can be used to understand the evolution of urban spatial structure.

See Laura Alessandretti et al., Marta C. González et al. and James Bagrow

This issue presents a Focus in which we highlight the potential of computational tools to help address sustainability and environmental issues — including those covered by the United Nations Sustainable Development Goals — as well as discuss how the tools themselves can be made more sustainable moving forward. The Focus consists of conversations with and insights from experts from various fields, addressing a wide range of applications under the umbrella of environmental and sustainable computing.

See Editorial

The cover depicts a computer-generated graphic of a lysine-specific molecular tweezer (metallic scaffolding) binding to a 14-3-3 protein from Homo sapiens (white mass). The protein's target lysine site is indicated by the glowing region. The approach introduced by Saldinger et al. in this issue was designed to accurately predict protein–nanoparticle interactions such as the one illustrated on the cover.

See Saldinger et al.

In this issue, we highlight two studies on moiré magnets. Li et al. developed a rotational and time-reversal equivariant neural network that can accurately model the dependence of the density functional theory Hamiltonian on atomic and magnetic superstructures. In another study, Yang et al. proposed a microscopic moiré spin model that enables the description of moiré magnetic exchange interactions via a sliding-mapping method. These methodological developments open opportunities for predicting emerging phenomena of magnetic superstructures, such as magnetic skyrmions. The cover image depicts — from top to bottom — magnetic field lines, magnetic configurations, a moiré lattice, Hamiltonian matrices, and neural networks.

See Editorial , Li et al. and Yang et al.

Complex materials with multiple elements have enabled various novel materials properties and applications. Insights from computational models can promote the effective exploration of vast chemical spaces resulting from such element coupling. In collaboration with Nature Materials, this issue features a Focus on complex element coupling, in which we at Nature Computational Science present a collection of expert opinions on the challenges and opportunities in model development that can further accelerate the rational design of such complex systems.

See Editorial

Innovative properties and desirable performance for materials design, especially associated with photonic applications, can be achieved via properly engineered disorder. In this issue, Sunkyu Yu develops the concept of an evolving scattering network to design disordered material systems with the properties of stealthy hyperuniformity, such as suppressing the scattering of waves across a select range of wavelengths. Light scattering is depicted on the cover.

See Sunkyu Yu and Yang Jiao

In this issue, Mingjian Wen, Kristin Persson and colleagues survey the different computational strategies available for chemical reaction network construction and analysis in a variety of applications, such as natural language processing and reaction property prediction. The opportunities for machine learning approaches, as well as the challenges that must still be overcome, are also discussed.

See Wen et al.