LARGE: AI language model maps antibiotic resistance genes in the aging gut
Peer-Reviewed Publication
SciOpen is an open access resource of scientific and technical content published by Tsinghua University Press and its publishing partners.
Below are the latest research news from journals hosted by SciOpen.
Updates every hour. Last Updated: 20-Oct-2026 21:16 ET (21-Oct-2026 01:16 GMT/UTC)
Antibiotic resistance is a growing global threat, yet how the collection of antibiotic resistance genes in the human gut—the antibiotic resistome—changes during aging has been poorly understood. Researchers at Westlake University have developed LARGE (https://github.com/hehang0618/LARGE), a large language model-based tool that detects and classifies antibiotic resistance genes with species-level resolution. Trained on 64,317 sequences, LARGE outperformed widely used tools such as DeepARG, RGI, PLM-ARG, and ARGNet, and identified more novel resistance genes than any comparator. Applying LARGE to more than 2000 gut metagenomes from three Chinese cohorts, the team found that resistance genes in 31 bacterial species changed significantly with age, and these changes were associated with 13 age-related chronic diseases, including chronic kidney disease. LARGE offers an online platform (https://large.lab.westlake.edu.cn), and all annotated data have been made publicly available.
Candidemia is a major cause of hospital-acquired bloodstream infections, with high mortality and substantial healthcare costs. However, delayed diagnosis, limited susceptibility testing, antifungal resistance, and the complexity of treatment continue to challenge effective patient management. This review examines recent advances in candidemia epidemiology, diagnosis, and therapy. It also discusses emerging approaches, including artificial intelligence-assisted diagnosis, vaccine-based prevention, and synergistic therapies, that could reshape candidemia management and improve outcomes.
A study of Scots pine plantations across Mediterranean Spain, temperate Romania and Northeast China found that climate constraints on growth vary by region. Soil moisture was the main limitation in dry Spanish sites, while temperature played a larger role in Romania and Northeast China. Spanish plantations showed sharper drought-related growth declines but stronger average recovery, although many did not recover within four years. The findings support region-specific forest management and improved models incorporating long-term drought, regional climate dynamics and local adaptation.
A new study in Forest Ecosystems tracked a 300-year temperate forest succession and found that particulate organic carbon (POC) declined at all depths, while mineral-associated organic carbon (MAOC) increased by 55% in surface soils but declined deeper down. These patterns were linked to shifts in microbial life-history strategies: surface MAOC accumulation associated with extracellular processing, while deeper carbon dynamics tied to microbial growth and declining necromass under resource scarcity. The findings suggest that forest age alone cannot predict carbon storage, soil depth and microbial mechanisms matter.
A study in Forest Ecosystems shows that aquaculture wastewater may weaken carbon stabilization and alter nitrogen regulation in mangrove soils. In China’s Maowei Sea, heavily affected soils contained much less dissolved organic carbon and substantially less organic carbon associated with crystalline iron minerals. Nutrient enrichment appeared to stimulate microbial carbon consumption and more reducing soil conditions, weakening iron-organic carbon coupling. At the same time, nitrogen cycling shifted toward greater ammonium retention and recycling rather than permanent removal. These changes could reduce the capacity of mangrove soils to preserve carbon over the long term and remove excess nitrogen, highlighting the importance of managing aquaculture discharge and monitoring carbon–iron-microbial interactions.