The paper presents an overview of Tahoe-x1 (Tx1), a new family of perturbation-trained single-cell foundation models developed by Tahoe Therapeutics, with up to three billion parameters. The authors explain that foundation models, which have transformed fields like computer vision, are now being adapted for single-cell biology, particularly for applications in precision oncology. Tx1 is pretrained on massive datasets, including the Tahoe-100M perturbation compendium, to learn generalized representations of genes, cells, and compounds. The model achieves state-of-the-art performance across multiple cancer-relevant benchmarks, such as predicting gene essentiality and modeling cellular responses to various chemical perturbations. Ultimately, the research emphasizes that scaling model capacity and training on large-scale perturbation data are crucial for developing robust, transferable models for cancer therapeutics.

References:

• Gandhi S, Javadi F, Svensson V, et al. Tahoe-x1: Scaling Perturbation-Trained Single-Cell Foundation Models to 3 Billion Parameters[J]. bioRxiv, 2025: 2025.10. 23.683759.

The review article introduces the concept of neural manifolds as a crucial framework for understanding how the brain generates animal behavior. It explains that although animal behavior relies on coordinated activity across vast neural populations, this activity exhibits a low-dimensional structure, which the authors posit is a result of these manifolds. Neural manifolds are defined as mathematical descriptions of the collective states a population of neurons can achieve, given constraints like connectivity and behavior, and the text explores the link between these manifolds and observed behavior. The authors further discuss various factors that shape neural manifolds across different timescales, such as learning and evolution, and propose future causal experimental approaches to validate and alter these structures to build a more comprehensive theory of brain function.

References:

• Perich M G, Narain D, Gallego J A. A neural manifold view of the brain[J]. Nature Neuroscience, 2025: 1-16.

The paper describes a study on developing synthetic receptors for enhancing CAR-T cell therapy against cancer. Researchers used a computational design platform to create allosteric switch receptors, named T-SenSERs, that respond to soluble factors in the tumor microenvironment (TME), specifically VEGFA or CSF1, with co-stimulation and cytokine signals. The goal is to overcome limitations of current CAR-T therapies, particularly in solid tumors, by providing context-dependent activation and boosting T cell persistence and anti-tumor function. Experimental validation in both lung cancer and multiple myeloma models demonstrated that combining CARs with these T-SenSERs (VMRFL and CMRFL) significantly enhanced anti-tumor activity in a ligand-dependent manner. This approach offers a novel framework for engineering cellular therapeutics with customized sensing and signaling responses.

References:

• Rath J A, Rudden L S P, Nouraee N, et al. Computational design of synthetic receptors with programmable signalling activity for enhanced cancer T cell therapy[J]. Nature Biomedical Engineering, 2025: 1-21.

The article explores the transformative potential of Artificial Intelligence (AI) in modernizing the study and application of Traditional Chinese Medicine (TCM) for treating stroke. The authors detail how TCM has shown efficacy in stroke recovery, but its complex, multi-component nature creates challenges for traditional research methods like compound screening and network pharmacology. The core of the paper focuses on how cutting-edge AI technologies—specifically Large Language Models (LLMs) and AI-integrated Virtual Cell (AIVC) models—can overcome these limitations by integrating diverse datasets, from ancient texts to multi-omics data, to elucidate multi-target therapeutic mechanisms. The article proposes theoretical frameworks for applying AI in both basic research (target prediction and validation) and clinical research (personalized treatment and diagnosis), aiming to move TCM from subjective, experience-based practice toward objective, precision medicine.

References:

• Meng Q, Mi Y, Wang F, et al. Intelligent technology leads the transformation of traditional Chinese medicine: Large models and virtual cells aid modern analysis of stroke treatment[J]. Pharmacological Research, 2025: 107953.

The article details a comprehensive study that utilized spatial and single-cell transcriptomics alongside imaging to map the cellular and molecular landscape of the developing human heart between 5.5 and 14 postconceptional weeks. Researchers created a high-resolution transcriptomic map, identifying 31 coarse-grained and 72 fine-grained cell states organized into functional niches, and specifically explored the formation of the cardiac pacemaker-conduction system (CPCS), heart valves, and autonomic innervation. A key finding was the identification of a resident neuroendocrine chromaffin cell population in the fetal heart, suggesting a human-specific role in coordinating the organ's response to hypoxia. The study also characterized the heterogeneity of endothelial and mesenchymal cells within the cardiac valves and septum, offering a valuable, open-access resource for investigating the genetic causes of heart disease.

References:

• Lázár E, Mauron R, Andrusivová Ž, et al. Spatiotemporal gene expression and cellular dynamics of the developing human heart[J]. Nature Genetics, 2025: 1-16.

This article details research on how low oxygen levels, or hypoxia, influence the development and differentiation of human lung epithelial cells, utilizing tissue-derived organoids. The authors found that hypoxia promotes airway differentiation in fetal lung progenitors while inhibiting alveolar cell fate, revealing that low oxygen acts as a developmental cue. Key to this process are the hypoxia-inducible factors (HIFs), particularly HIF1α and HIF2α, which have distinct roles: HIF1α is necessary for airway differentiation, whereas HIF2α specifically promotes basal cell fate but suppresses other airway and alveolar fates. Furthermore, the study indicates that chronic hypoxia can convert differentiated alveolar type 2 (AT2) cells, even those from adult lungs, into aberrant airway-like cells, suggesting a role for hypoxia as a pathological factor in lung diseases like fibrosis.

References:

• Dong Z, Wit N, Agarwal A, et al. Hypoxia promotes airway differentiation in the human lung epithelium[J]. Cell Stem Cell, 2025.

This article details a comprehensive spatial immune profiling study focusing on adult-type diffuse gliomas, which are often considered immunologically "cold" tumors. Researchers identified tertiary lymphoid structures (TLSs) in 15% of these gliomas and further characterized them into three distinct subtypes: T-low-TLS, B:T-TLS, and PC-TLS (plasma cell rich). The study demonstrates that functional TLS maturation in gliomas, particularly the B:T-TLS and PC-TLS subtypes, involves clonal lymphocyte expansion, plasma cell formation, and interactions between cytotoxic T cells and dendritic cells. Crucially, the presence of these functional TLSs is strongly associated with improved overall patient survival, suggesting they represent an immune-permissive tumor microenvironment and may serve as a critical prognostic and predictive biomarker for immunotherapy susceptibility in a subset of glioma patients.

References:

• Cakmak P, Lun J H, Singh A, et al. Spatial immune profiling defines a subset of human gliomas with functional tertiary lymphoid structures[J]. Immunity, 2025.

The article investigats the role of astrocytes at the border of a central nervous system injury, specifically after a stroke. The key finding is that interferon-responsive astrocytes (IRRAs), identified by high expression of the protein BST2 (bone marrow stromal cell antigen 2), accumulate at the lesion border and drive microglial recruitment and activation. Mechanistically, BST2 promotes the activation of the NF-κB pathway via PKCβII phosphorylation, leading to the secretion of C3, which subsequently engages the microglial C3aR to mediate harmful astrocyte-microglia crosstalk. Genetic or antibody blockade of BST2 reduces these interactions, diminishing microglial activation and improving early motor function and neuronal survival, thus identifying BST2 as a potential therapeutic target for stroke recovery.

References:

• Zhang S, Yuan M, Zhou J, et al. BST2 expression at astrocyte borders promotes microglial recruitment via the C3/C3aR signaling[J]. Neuron, 2025.

This article reports on a scientific study demonstrating that oral vitamin C supplementation delays ovarian aging in non-human primates. The authors conducted a multi-year experiment in cynomolgus macaques, finding that vitamin C diminished common aging markers such as oxidative stress and follicular depletion. Utilizing advanced single-cell transcriptomic analysis, the research revealed that vitamin C had a rejuvenating effect on both oocytes and somatic cells, effectively reducing their predicted biological age. Mechanistically, the study suggests that these geroprotective benefits are partly mediated by the activation of the NRF2 antioxidant pathway, positioning vitamin C as a potentially translatable strategy for combating reproductive aging.

References:

• Jing Y, Lu H, Li J, et al. Vitamin C conveys geroprotection on primate ovaries[J]. Cell Stem Cell, 2025.

This article introduces a novel human heart-macrophage assembloid (hHMA) platform, which integrates pluripotent stem cell-derived embryonic monocytes into heart organoids to mimic natural immune-cardiac interactions during development. The researchers demonstrate that these tissue-resident macrophages (TRMPs) successfully integrate, persist long-term, and regulate vital cardiac functions, including morphogenesis, efferocytosis, and electrical conduction. Furthermore, the hHMAs provide a powerful in vitro model for disease by demonstrating that chronic activation of the NLRP3 inflammasome in the assembloids induces arrhythmogenic features consistent with atrial fibrillation (AF). The study uses various advanced techniques like single-cell transcriptomics and proteomics to characterize the TRMPs and their role in modulating cell signaling, including the release of SPP1-positive extracellular vesicles (EVs). Overall, the work establishes a sophisticated platform for studying human heart development and inflammation-driven arrhythmias, offering potential for future mechanistic and preclinical drug screening.

References:

• O’Hern C, Caywood S, Aminova S, et al. Human heart-macrophage assembloids mimic immune-cardiac interactions and enable arrhythmia disease modeling[J]. Cell Stem Cell, 2025.