Research articles

Realistic quantum mechanical simulations are computationally costly to perform but can be approximated using neural network models. Li and colleagues propose a forward propagation method in lieu of traditional backpropagation to speed up these neural network-based approaches.

Great advances in protein structure prediction have been made with recent deep learning-based methods, but proteins interact with their environment and can change shape drastically when binding to ligand molecules. To predict the 3D structure of these combined protein–ligand complexes, Qiao et al. developed a generative diffusion model with biophysical constraints and geometric deep learning.

AI-enabled diagnostic applications in healthcare can be powerful, but study design is very important to avoid subtle issues of bias in the dataset and evaluation. Coppock et al. demonstrate how an AI-based classifier for diagnosing SARS-Cov-2 infection from audio recordings can seem to make predictions with high accuracy but shows much lower performance after taking into account confounders, providing insights in study design and replicability in AI-based audio analysis.

Machine learning techniques are widely employed in chemical science, but are application specific and their development requires dedicated expertise. Jablonka and colleagues fine-tune the GPT-3 model and show that it can provide surprisingly accurate answers to a wide range of chemical questions.

Recent years have seen many advances in deep learning models for protein design, usually involving a large amount of training data. Focusing on potential clinical impact, Garton et al. develop a variational autoencoder approach trained on sparse data of natural sequences of adenoviruses to generate large proteins that can be used as viral vectors in gene therapy.

Designing antibodies and assessing their biophysical properties for potential therapeutic development is challenging with current computational methods. Ramon et al. have developed a deep learning approach called AbNatiV, based on a vector-quantized variational encoder that accurately assesses the nativeness of antibodies and nanobodies, which are small single-domain antibodies that have recently attracted considerable interest.

Drug design has recently seen immense improvements in computational methods, but models can still struggle generalizing across binding pockets. Feng and colleagues combine a language model with geometric deep learning to provide efficient generation of potential new drugs.

Accurate real-time tracking of dexterous hand movements and interactions has applications in human–computer interaction, the metaverse, robotics and tele-health. Capturing realistic hand movements is challenging due to the large number of articulations and degrees of freedom. Tashakori and colleagues report accurate and dynamic tracking of articulated hand and finger movements using machine-learning powered stretchable, washable smart gloves.

Magnetic microrobots are of considerable interest for non-invasive biomedical applications but it is challenging to develop a general strategy for controlling microrobot positions, for varying configurations and environments. Choi et al. develop a reinforcement learning control method, training the model in a simulation environment for initial exploration after which the learning process is transferred to a physical electromagnetic actuation system.

Multi-animal behaviour quantification is pivotal for deciphering animal social behaviours and has broad applications in neuroscience and ecology. Han and colleagues develop a few-shot learning framework for multi-animal 3D pose estimation, identity recognition and social behaviour classification.

Feed-forward neural networks have become powerful tools in machine learning, but their behaviour during optimization is still not well understood. Ciceri and colleagues find that during optimization, class representations first separate and then rejoin, prompted by specific elements of the training set.

Machine learning methods in cheminformatics have made great progress in using chemical structures of molecules, but a large portion of textual information remains scarcely explored. Liu and colleagues trained MoleculeSTM, a foundation model that aligns the structure and text modalities through contrastive learning, and show its utility on the downstream tasks of structure–text retrieval, text-guided editing and molecular property prediction.

Theoretical frameworks aiming to understand deep learning rely on a so-called infinite-width limit, in which the ratio between the width of hidden layers and the training set size goes to zero. Pacelli and colleagues go beyond this restrictive framework by computing the partition function and generalization properties of fully connected, nonlinear neural networks, both with one and with multiple hidden layers, for the practically more relevant scenario in which the above ratio is finite and arbitrary.

Interest in using large language models such as ChatGPT has grown rapidly, but concerns about safe and responsible use have emerged, in part because adversarial prompts can bypass existing safeguards with so-called jailbreak attacks. Wu et al. build a dataset of various types of jailbreak attack prompt and demonstrate a simple but effective technique to counter these attacks by encapsulating users’ prompts in another standard prompt that reminds ChatGPT to respond responsibly.

Machine learning models have been widely used in the inverse design of new materials, but typically only linear properties could be targeted. Bastek and Kochmann show that video diffusion generative models can produce the nonlinear deformation and stress response of cellular materials under large-scale compression.

Virtual drug design has seen recent progress in methods that can generate new molecules with specific properties. Separately, methods have also improved in the task of computationally predicting the outcome of chemical reactions. Qiang and colleagues use the close relation of the two problems to train a model that aims at solving both tasks.

Data-driven surrogate models are used in computational physics and engineering to greatly speed up evaluations of the properties of partial differential equations, but they come with a heavy computational cost associated with training. Pestourie et al. combine a low-fidelity physics model with a generative deep neural network and demonstrate improved accuracy–cost trade-offs compared with standard deep neural networks and high-fidelity numerical solvers.

Single-cell transcriptomics has provided a powerful approach to investigate cellular properties at unprecedented resolution. Sha et al. have developed an optimal transport-based algorithm called TIGON that can connect transcriptomic snapshots from different time points to obtain collective dynamical information, including cell population growth and the underlying gene regulatory network.

A fundamental question in neuroscience is what are the constraints that shape the structural and functional organization of the brain. By bringing biological cost constraints into the optimization process of artificial neural networks, Achterberg, Akarca and colleagues uncover the joint principle underlying a large set of neuroscientific findings.

Deep learning is a powerful method to process large datasets, and shown to be useful in many scientific fields, but models are highly parameterized and there are often challenges in interpretation and generalization. David Gleich and colleagues develop a method rooted in computational topology, starting with a graph-based topological representation of the data, to help assess and diagnose predictions from deep learning and other complex prediction methods.