Skip to main content
Search

Chinese Academy of Medical Sciences & Peking Union Medical College

6 related articles

Subscribe to updates
  1. arXiv

    Pathology agents stop answering one slide at a time: PlanPath turns a question into an exploration plan and reaches 72.71% total accuracy on renal cancer

    A pathology case often spans dozens of slides from different specimens, yet existing pathology agents analyze one slide at a time and describe morphology rather than measure it. PlanPath turns a diagnostic question into an exploration plan, routes to the slides likely to carry the evidence, then zooms from tissue architecture down to cells for counts and proportions, aggregating cross-slide evidence into a traceable answer. On patient-level renal and gastric cancer datasets, PlanPath reaches total accuracy of 72.71% and 60.24%, 12.14 and 9.43 percentage points above the strongest baseline.

  2. Skin microbiome catalog fills a reference-genome gap: 452,000 genomes, most species previously unrepresented

    Reference genomes for the human skin microbiome have been scarce, especially outside bacteria and across global strain variation. The Global Skin Metagenomic Data resource combined 12,409 newly sequenced and 4,396 public skin metagenomes with 3,077 isolate genomes to reconstruct 452,397 metagenome-assembled genomes and add 42,758 previously uncharacterized taxa. From 1.71 billion predicted protein sequences it consolidated 1,272,886 protein families, 79,434 of which had no homology hit in Pfam or RefSeq.

  3. arXiv

    A coronary agent now shows its evidence before its answer: the gate passes 76.7%, and rejects half of reference lesions

    Panvascular intervention research is starting to write traceability into an agent's output contract: ScienceBuddy-PV fixes a shared convolutional model on the public ARCADE coronary angiography benchmark and evaluates only an evidence gate and action audit, where the gate accepts 76.7% of stenosis-frame candidates against 100% for a score-only comparator. On the locked test set, stenosis Dice is 0.231 and AUROC is 0.957 at a pixel prevalence of 0.00975, while reference lesion candidates are accepted 44.3% and rejected 55.7% of the time. The result is an image-level technical validation; recursive harness optimization and model learning remain protocols rather than executed experiments.

  4. Diabetes

    Metabolomic clustering of 24,638 UK Biobank participants with prediabetes yields three subtypes with progressively higher type 2 diabetes, cardiovascular, and chronic kidney disease risk

    In 24,638 UK Biobank participants with prediabetes, LASSO and elastic net selection of nuclear magnetic resonance metabolomic markers yielded 16 biomarkers, and k-means clustering in the training set (12,322 participants) identified low-, intermediate-, and high-metabolic-risk subtypes that showed progressively higher risks of type 2 diabetes, cardiovascular disease, and chronic kidney disease in the validation set (12,316 participants), with diet-quality associations differing across subtypes and Mendelian randomization suggesting potential causal links for several metabolomic biomarkers.

  5. Research Square

    Clinical quality of large language model answers to questions about pancreatic cancer

    This study had eight large language models (ChatGPT-5.5 Instant, Claude Sonnet 4.6, Gemini 3 Flash, Grok 4.20, Microsoft Copilot, Doubao Seed 1.8, DeepSeek-V4, and Qwen 3.6) answer ten patient-oriented pancreatic cancer questions on three consecutive days, with five evaluators blinded to platform identity rating all 240 responses across six dimensions of clinical quality (1,200 ratings), finding that overall performance differed among platforms, with Grok 4.20 and DeepSeek-V4 having the highest mean scores and Doubao Seed 1.8 the lowest, differences being greatest in actionability and specificity, adjacent-day score fluctuation being most pronounced for Doubao Seed 1.8 and Qwen 3.

  6. bioRxiv

    Deep reinforcement learning-driven discovery of a MsbA-targeted small-molecule antibiotic for the treatment of Acinetobacter baumannii infection

    Using cerastecin Cpd 4 as a template, this study applied two AI tools, Link-INVENT and AutoMolDesigner, for molecular design and chemical derivatization, leading to the discovery of the MsbA-targeted small molecule Y-11 with an MIC of 0.5 g/mL against A. baumannii, equivalent potency to Cpd4 against carbapenem-resistant A. baumannii, lower cytotoxicity, hemolysis, and spontaneous resistance frequency, effective reduction of bacterial loads in infected mice, and a proposed mechanism in which Y-11 inhibits lipooligosaccharide transport and impairs outer membrane formation, probably by competitively binding the substrate binding site of MsbA and modulating ATPase activity.

You’re all caught up