Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript.
This paper introduces an algorithm to uncover laws of skill acquisition from naturally occurring data. By combining deep learning and symbolic regression, it accurately identifies cognitive states and extracts algebraic equations.
The authors develop the tool RESHAPE to share reference panels in a safer way. The genome–phenome links in reference panels can generate re-identification threats and RESHAPE breaks these links by shuffling haplotypes while preserving imputation accuracy.
A method based on a vector-quantized variational autoencoder, called CASTLE, can interpretably extract discrete latent embeddings and quantitatively generate the cell-type-specific feature spectrum for single-cell chromatin accessibility sequencing data.
Cooperation is not merely a dyadic phenomenon, it also includes multi-way social interactions. A mathematical framework is developed to study how the structure of higher-order interactions influences cooperative behavior.
This study introduces SANGO, a method for accurate single-cell annotation leveraging genomic sequences around accessibility peaks within single-cell ATAC sequencing data. SANGO consistently outperforms existing methods across diverse datasets for identification of cell type and detection of unknown tumor cells. SANGO enables the discovery of cell-type-specific functional insights through expression enrichment, cis-regulatory chromatin interactions and motif enrichment analyses.
A method is developed for the directional optimization of multiple properties without prior knowledge on their nature. Using a large ligand dataset, diverse metal complexes are found along the Pareto front of vast chemical spaces.
Andre Berndt and colleagues introduce a machine learning approach to enhance the biophysical characteristics of genetically encoded fluorescent indicators, deriving and testing in vitro new GCaMP mutations that surpass the performance of existing fast GCaMP indicators.
M-OFDFT is a deep learning implementation of orbital-free density functional theory (OFDFT) that achieves DFT-level accuracy on molecular systems with lower cost complexity, and can extrapolate to much larger molecules than those seen during training.
An optimization algorithm is used to discover guest molecules based on knowing only the structure of the host. The molecules are represented as 3D volumes, optimized to improve host–guest interaction and converted into SMILES using a transformer model.
SCORPION is an algorithm to model gene regulatory networks based on single-cell data. The authors show that SCORPION outperforms other methods, accurately detects transcription factor activity and can potentially help with the discovery of disease markers.
Automated algorithm discovery has been difficult for artificial intelligence given the immense search space of possible functions. Here explainable neural networks are used to discover algorithms that outperform those designed by humans.
DNA microscopy reconstructs the spatial organization of a sample from a neighborhood graph. In this work, MinIPath efficiently corrects errors from these graphs that distort the reconstruction, both in simulated and experimental data.
The authors develop a general method that combines machine learning and physics to construct macroscopic dynamics directly from microscopic observations, leading to an intuitive understanding of polymer stretching in elongational flow.
EmerGNN, a method to predict interactions for emerging drugs, may improve patient care and drug development by providing insight into the effects of using biomedical networks in interaction predictions.
Using registry data from Denmark, Lehmann et al. create individual-level trajectories of events related to health, education, occupation, income and address, and also apply transformer models to build rich embeddings of life-events and to predict outcomes ranging from time of death to personality.
A diffusion model that generates chemical reactions in 3D with all desired symmetries preserved is established and shown to reduce transition state search from days to seconds and complement intuition-based reaction exploration with generative AI.
Developing predictive mechanistic models in biology is challenging. Elektrum uses neural architecture search, kinetic models and transfer learning to discover CRISPR–Cas9 cleavage kinetics, achieving high performance and biophysical interpretability.
Signal peptides (SPs) are vital for protein–transmembrane communication. In this work, the authors introduce USPNet, a deep learning method based on a protein language model for SP prediction that shows both high sensitivity and efficiency, thereby contributing to the identification of novel SPs.
SRDTrans is a self-supervised denoising method for fluorescence images powered by spatial redundancy sampling and a dedicated transformer network that achieves good performance on fast dynamics and various imaging modalities.
VSSR-MC is a Markov chain method based on virtual adsorption sites that interfaces with a neural network force field to provide fast, accurate and comprehensive sampling of material surfaces.