Summary of Nature publications, focusing on contents relevant to AI and life sciences
Africa South of the Sahara, Mining, Conservation of Natural Resources, Forests, Spatio-Temporal Analysis
Researchers quantified how mining drives forest loss in sub-Saharan Africa by analyzing 16,627 mines from 2001 to 2020 with satellite data and a difference-in-differences approach. They found direct mining operations (pits, tailings, spoil heaps) caused 187,000 hectares of deforestation. More striking, mining triggered an additional 8 percentage points of deforestation within 1 km compared to unmined areas, and elevated losses up to 20 km away persisted for a decade. For every hectare cleared directly by a mine, an average of 34 additional hectares were lost within five years via indirect effects like new farms and settlements. Cobalt and copper mines, crucial for energy transition technologies, caused the most indirect forest loss. The study urges incorporating offsite deforestation into environmental impact assessments to achieve zero-deforestation supply chains for critical minerals.
Saccharum, Polyploidy, Genome, Plant, Genomics, Sucrose, Haplotypes, Genome-Wide Association Study, Chromosomes, Plant, Alleles, Genes, Plant, Quantitative Trait Loci, Phenotype
Researchers cracked the genetic code behind sugarcane's massive sugar production by assembling a completely phased genome of POJ2878, the foundational cultivar of modern sugarcane. The assembly revealed 118 chromosomes and extensive recombination between subgenomes. By comparing alleles and expression across 981 accessions, they traced POJ2878's widespread influence and found key selection sweeps tied to cold tolerance, cell size, and tillering. Their polyploid-aware genome-wide association study pinpointed loci linked to sugar storage and cell development, including a functionally verified sucrose transporter. This work lays a critical genomic groundwork for faster, targeted breeding of sugarcane—and potentially other complex polyploid crops—to support global food and bioenergy needs.
Animals, Humans, Artificial Intelligence, Computational Biology, COVID-19, Empirical Research, Forecasting, Large Language Models, SARS-CoV-2, Single-Cell Analysis, Software, Zebrafish
Researchers built an AI system, Empirical Research Assistance (ERA), that generates scientific software aimed at optimizing a defined quality metric. ERA combines a large language model with tree search to systematically explore and refine software solutions, including pulling in complex ideas from external literature. Across multiple domains, ERA produced competitive results: in single-cell bioinformatics it devised 40 methods surpassing top human approaches on a public leaderboard; for COVID-19 hospitalization forecasting it generated 14 models beating CDC ensemble and individual baselines. It also created expert-level code for geospatial tasks, zebrafish neural activity prediction, numerical integration, and a novel rule-based time-series construction. The authors argue this shows AI can autonomously develop new, effective computational methods across diverse sciences, potentially accelerating the research cycle by removing a key bottleneck in experiment-driven discovery.
Humans, Triple Negative Breast Neoplasms, Female, Tumor Microenvironment, Neoadjuvant Therapy, Interferons, Macrophages, Single-Cell Analysis, Single-Cell Gene Expression Analysis, Stromal Cells, Gene Expression Regulation, Neoplastic, Spatial Transcriptomics, Cell Cycle, Treatment Outcome
Researchers profiled nearly 430,000 single cells from tumor samples of 101 patients with untreated triple-negative breast cancer, along with spatial transcriptomics from 44 patients, before neoadjuvant chemotherapy. They grouped tumors into four patient-level archetypes based on cancer-cell gene expression and identified 13 metaprograms capturing intra-tumoral variation. The tumor microenvironment included 49 immune and stromal cell states, many altered relative to normal breast tissue. By mapping how cancer cells co-occur with these cell types spatially, they defined eight cellular communities, or ecotypes. Notably, the key predictors of good chemotherapy response were macrophage subtypes and specific cancer-cell programs involving interferon signaling, HLA expression, and cell-cycle activity—contrasting with earlier emphasis on T cells. These findings clarify how untreated TNBCs are organized and point to new biomarkers or targets linked to treatment outcomes.
Electron Wigner solids are a testbed for how electron-electron interactions compete with disorder, a longstanding puzzle in condensed matter physics. Progress stalled because experiments lacked single-defect resolution and theories couldn't match real data. This study combined scanning tunneling microscopy with neural-quantum-state quantum Monte Carlo simulations to study disordered 2D Wigner solids in bilayer MoSe₂ devices, using realistic disorder maps from experiments in the simulations. The team found two distinct regimes: with low short-range disorder, long-range disorder dominates, producing mixed solid-liquid phases, local re-entrant melting/crystallization, and Friedel oscillations. With high short-range disorder, these effects vanish, yielding a more robust amorphous Wigner solid that survives at higher electron densities. The work presents a new combined experimental-theoretical framework for probing disordered quantum solids at the single-defect level.
Diffusion, Friction, Hydrogen Bonding, Light, Molecular Dynamics Simulation, Nanotubes, Carbon, Quantum Mechanics, Quantum Theory, Water
Researchers found that shining near-infrared light on water-dispersed carbon nanotubes causes a quantum-level friction effect that slows their Brownian motion by about half. This friction is triggered by excitons—bound electron-hole pairs—whose fluctuating electric fields drag against water molecules. The effect vanishes when excitons are pinned at engineered defect sites, and chemically boosting or suppressing nanotube fluorescence correspondingly speeds up or slows down diffusion up to two-fold. Ultrafast terahertz spectroscopy revealed an instant coupling to water's Debye modes, followed by a slower heating-like response in the hydrogen-bond network. Molecular dynamics simulations confirmed that exciton dipole fluctuations generate the frictional force. The work demonstrates that electronic excitations can be harnessed to tune nanoscale motion and fluid behavior, opening possibilities for light-controlled nanofluidics or mechanical actuators in aqueous environments.
Humans, DNA Methylation, Immunotherapy, Melanoma, Spatial Transcriptomics, Tumor Microenvironment, Carcinoma
The researchers developed a machine-learning framework to map recurring multicellular organizations in the tumor microenvironment, which they call spatial ecotypes (SEs). Using over ten million spatial transcriptomic measurements from multiple human cancer types, they identified nine conserved SEs, each with distinct biological features, spatial patterns, and links to clinical outcomes, including immunotherapy response. Notably, these SEs could be detected indirectly from DNA methylation data and, remarkably, from plasma cell-free DNA via deep learning. In nearly one hundred melanoma patients, the relative abundance of SEs in blood strongly predicted how well patients responded to immunotherapy. This work defines fundamental building blocks of tumor organization and offers a non-invasive, liquid-biopsy-compatible way to monitor the tumor microenvironment, potentially improving risk assessment and enabling more personalized cancer treatment.
Protein Folding, Models, Molecular, Proteins, Protein Domains, Protein Stability, Thermodynamics, Hydrogen Deuterium Exchange-Mass Spectrometry, Nuclear Magnetic Resonance, Biomolecular, Protein Conformation, Mutation, Deuterium Exchange Measurement, Protein Structure, Secondary, Mass Spectrometry
Most studies of proteins focus on their stable, folded shapes, but proteins also constantly visit rare, higher-energy conformations that are crucial for function and disease. These fleeting states have been hard to measure. Researchers developed a high-throughput method using hydrogen–deuterium exchange mass spectrometry to probe the energy of these conformational fluctuations in 5,778 small protein domains simultaneously. They found that energetically similar sequences can hide very different fluctuation patterns. Detailed NMR on 13 domains revealed that fluctuations often involve entire secondary structure elements that are less stable than the protein as a whole. Machine learning on the data linked specific structural features to experimentally measured instability, and the team used this to engineer mutations that stabilized weak regions. The work provides a large dataset for training models to predict protein energy landscapes and offers a scalable way to chart these dynamic behaviors, potentially improving protein design and understanding of misfolding diseases.
Female, Humans, Male, Demography, European People, Genetics, Population, Genome, Human, Germany, History, Ancient, Pedigree, Roman World
Researchers sequenced 258 ancient genomes from southern Germany's former Roman frontier and compared them with thousands of other ancient and modern genomes to investigate how this region's population and society transformed between 400 and 700 CE. They found a major demographic shift around the late fifth century, when a northern European founding population mixed extensively with genetically diverse Roman provincial groups, eventually producing a population genetically similar to modern Central Europeans by the early seventh century. By reconstructing family trees and using newly developed analytical methods, they estimated a generation time of 28 years, life expectancies of about 40 years for women and 43 for men, high infant mortality, and a social structure centered on nuclear families with lifelong monogamy, incest avoidance, and flexible lineage continuation. The findings suggest small-group mobility and intermarriage, not large-scale migration, reshaped this frontier society, with family practices showing continuity from Late Roman times.
Comments (0)
No comments yet. Log in to leave a comment.
You must be logged in to leave a comment.
Login to Comment