Browse Articles

Spatial mapping of innate lymphoid cells in human lymphoid tissues and lymphoma at single-cell resolution

Autologous HIV-specific T cell therapy targeting conserved epitopes is well-tolerated in six adults with HIV: an open-label, single-arm phase 1 study

Combining cross-sectional and longitudinal genomic approaches to identify determinants of cognitive and physical decline

Abstract Large-scale genomic studies focusing on the genetic contribution to human aging have mostly relied on cross-sectional data. With the release of longitudinally curated aging phenotypes by the UK Biobank (UKBB), it is now possible to study aging over time at genome-wide scale. In this work, we evaluated the suitability of competing models of change in realistic simulation settings, performed genome-wide association scans on simulation-validated measures of age-related deweekcline, and followed up with LD-score regression and Mendelian Randomization (MR) analyses. Focusing on global cognitive and physical function, we observed marked differences between baseline function (θ) and accelerated decline (Δ). Both outcomes showed distinct heritability levels (e.g., 31.38% $${h}_{\theta }^{2}$$ h θ 2 versus 3.15% $${h}_{\Delta }^{2}$$ h Δ 2 for physical function) and different associated loci (e.g., DUSP6 specific to physical Δ). Further, we found little commonalities across the two dimensions of aging—while cognitive decline was largely driven by Alzheimer’s disease liability (standardized MR-effect, γ = 0.17), physical decline was mostly impacted by telomere length (γ = −0.05) and bone mineral density (γ = −0.05). Our work highlights the utility of longitudinal genomic efforts to scrutinize age-dependent genetic and environmental effects on physical and cognitive outcomes. Careful modelling and attention to participation characteristics are, however, crucial for valid inference.

PRMT3 reverses HIV-1 latency by increasing chromatin accessibility to form a TEAD4-P-TEFb-containing transcriptional hub

Long-Term Efficacy of Pembrolizumab and the Clinical Utility of ctDNA in Locally Advanced dMMR/MSI-H Solid Tumors

Blood collection tube and RNA purification method recommendations for extracellular RNA transcriptome profiling

A human brain map of mitochondrial respiratory capacity and diversity

Blood of man who’s had 200 snake bites helps make a potent antivenom

AI scientist ‘team’ joins the search for extraterrestrial life

Trump proposes unprecedented budget cuts to US science

Emergence of accurate atomic energies from machine-learned noble-gas potentials

The quantum theory of atoms in molecules gives access to well-defined local atomic energies. Due to their locality, these energies are potentially interesting in fitting atomistic machine learning models as they inform about physically relevant properties. However, computationally, quantum-mechanically accurate local energies are notoriously difficult to obtain for large systems. Here, we show that by employing semiempirical correlations between different components of the total energy, we can obtain well-defined local energies at a moderate cost. We employ this methodology to investigate energetics in noble liquids or argon, krypton, and their mixture. Instead of using these local energies to fit atomistic models, we show how well these local energies are reproduced by machine-learned models trained on the total energies. The results of our investigation suggest that smaller neural networks, trained only on the total energy of an atomistic system, are more likely to reproduce the underlying local energy partitioning faithfully than larger networks. Furthermore, we demonstrate that networks more capable of this energy decomposition are, in turn, capable of transferring to previously unseen systems. Our results are a step toward understanding how much physics can be learned by neural networks and where this can be applied, particularly how a better understanding of physics aids in the transferability of these neural networks.

Identification of undetected SARS-CoV-2 infections by clustering of Nucleocapsid antibody trajectories

Abstract During the COVID-19 pandemic, numerous SARS-CoV-2 infections remained undetected. We combined results from routine monthly nose and throat swabs, and self-reported positive swab tests, from a UK household survey, linked to national swab testing programme data from England and Wales, together with Nucleocapsid (N-)antibody trajectories clustered using a longitudinal variation of K-means (N = 185,646) to estimate the number of infections undetected by either approach. Using N-antibody (hypothetical) infections and swab-positivity, we estimated that 7.4% (95%CI: 7.0–7.8%) of all true infections (detected and undetected) were undetected by both approaches, 25.8% (25.5–26.1%) by swab-positivity-only and 28.6% (28.4–28.9%) by trajectory-based N-antibody-classifications-only. Congruence with swab-positivity was respectively much poorer and slightly better with N-antibody classifications based on fixed thresholds or fourfold increases. Using multivariable logistic regression N-antibody seroconversion was more likely as age increased between 30–60 years, in non-white participants, those less (recently/frequently) vaccinated, for lower cycle threshold values in the range above 30, and in symptomatic and Delta (vs. BA.1) infections. Comparing swab-positivity data sources showed that routine monthly swabs were insufficient to detect infections and incorporating national testing programme/self-reported data substantially increased detection. Overall, whilst N-antibody serosurveillance can identify infections undetected by swab-positivity, optimal use requires fourfold-increase-based or trajectory-based analysis.

Advanced internet of things enhanced activity recognition for disability people using deep learning model with nature-inspired optimization algorithms

apoCHARMM: High-performance molecular dynamics simulations on GPUs for advanced simulation methods

We present apoCHARMM, a high-performance molecular dynamics (MD) engine optimized for graphics processing unit (GPU) architectures, designed to accelerate the simulation of complex molecular systems. The distinctive features of apoCHARMM include single-GPU support for multiple Hamiltonians, computation of a full virial tensor for each Hamiltonian, and full support for orthorhombic periodic systems in both P1 and P21 space groups. Multiple Hamiltonians on a single GPU permit rapid single-GPU multi-dimensional replica exchange methods, multi-state enveloping distribution sampling methods, and several efficient free energy methods where efficiency is gained by eliminating post-processing requirements. The combination of these capabilities enables constant-pH molecular dynamics in explicit solvent with enveloping distribution sampling, where Hamiltonian replica exchange can be performed on a single GPU with minimal host-GPU memory transfers. A full atomic virial tensor allows support for many different pressure, surface tension, and temperature ensembles. Support for orthorhombic P21 systems allows for the simulation of lipid bilayers, where the two leaflets have equalized chemical potentials. apoCHARMM uses CUDA and modern C++ to enable efficient computation of energy, force, restraint, constraint, and integration calculations directly on the GPU. This GPU-exclusive design focus minimizes host-GPU memory transfers, ensuring optimal performance during simulations, with such transfers occurring only during logging or trajectory saving. Benchmark tests demonstrate that apoCHARMM achieves competitive or superior performance when compared to other GPU-based MD engines, positioning it as a versatile and useful tool for the molecular dynamics community.

Structural and molecular basis of PCNA-activated FAN1 nuclease function in DNA repair

Genetic diversity of A(H5N1) avian influenza viruses isolated from birds and seals in Russia in 2023

Assessment of RPA and <i>σ</i>-functional methods for the calculation of dipole moments and static polarizabilities and hyperpolarizabilities

In the present paper, we assess the performance of methods based on the random phase approximation (RPA) and on σ-functionals for predicting static optical properties, i.e., dipole moment, polarizability, and first and second hyperpolarizability, of small- and medium-sized molecules, including chain-like systems. First, we provide accurate reference data by coupled-cluster singles, doubles, with perturbative triples calculations with sufficiently large basis sets. The RPA and σ-functional calculations are carried out post-self-consistently using input orbitals and eigenvalues from the hybrid density-functional calculation. The optimal fraction of exact non-local exchange in these calculations is found to be quite high, around 0.5–0.6 in RPA and around 0.8–1.0 in σ-functional methods. σ-functional methods, however, proved to be less sensitive than RPA methods with respect to the amount of exact non-local exchange used in the generation of their input data. σ-functional methods are shown to outperform in accuracy RPA methods and various other considered density-functional theory methods for static optical properties and, thus, are well-suited for the calculation of linear and non-linear optical properties.

Hierarchical glycolytic pathways control the carbohydrate utilization regulator in human gut Bacteroides

Klotho plasma levels are an independent predictor of mortality in women with acute coronary syndrome

GPU acceleration of hybrid functional calculations in the SPARC electronic structure code

We present a Graphics Processing Unit (GPU)-accelerated version of the real-space SPARC electronic structure code for performing hybrid functional calculations in generalized Kohn–Sham density functional theory. In particular, we develop a batch variant of the recently formulated Kronecker product-based linear solver for the simultaneous solution of multiple linear systems. We then develop a modular, math kernel based implementation for hybrid functionals on NVIDIA architectures, where computationally intensive operations are offloaded to the GPUs, while the remaining workload is handled by the central processing units (CPUs). Considering bulk and slab examples, we demonstrate that GPUs enable up to 8× speedup in node-hours and 80× in core-hours compared to CPU-only execution, reducing the time to solution on V100 GPUs to around 300 s for a metallic system with over 6000 electrons, and significantly reducing the computational resources required for a given wall time.