Citations

Abstract Immune checkpoint inhibitors (ICIs) can produce durable tumor control, yet many patients experience primary or acquired resistance, mixed lesion-level responses, or immune-related adverse events. Tissue and blood biomarkers provide biologically informative but spatially incomplete views of a heterogeneous and time-varying tumor–host system. Radiologic imaging complements these assays by repeatedly measuring whole-body tumor burden, organ-specific disease patterns, lesion-level heterogeneity, and treatment-related change. This critical narrative review evaluates imaging-centered artificial intelligence (AI) for precision cancer immunotherapy, including baseline and serial CT, MRI, and PET; radiomics and learned image representations; lesion-level and whole-body heterogeneity; toxicity assessment; and integration with pathology, genomics, circulating tumor DNA, laboratory measurements, and electronic health records. Representative cohorts range from early radiogenomic studies of approximately 100–200 patients to externally validated and longitudinal models involving hundreds or thousands of patients. Rather than treating all reported performance estimates as equally actionable, we appraise the clinical question, cohort design, endpoint, validation strategy, calibration, and intended clinical action. Current evidence supports noninvasive immune phenotyping, risk stratification, longitudinal monitoring, and multimodal biomarker integration. However, most models predict outcomes among patients already receiving ICIs and therefore remain prognostic under treatment rather than predictive of differential treatment benefit. Radiotherapy (RT) combined with ICI serves as a focused imaging use case because lesion identity, dose distribution, fractionation, sequence, and thoracic and normal-tissue exposure must be linked to defined actions. Existing patient-level AI studies estimate efficacy or toxicity after treatment selection; none has prospectively shown that an AI-selected RT strategy improves patient outcomes. Translation will require patient-level leakage control, calibration, geographic and temporal validation, robustness to missing modalities, and prospective evaluation of prespecified model-guided decisions.

Abstract With the rapid advancements in large language model technology and the emergence of bioinformatics‐specific language models (BioLMs), there is a growing need for a comprehensive analysis of the current landscape, computational characteristics, and diverse applications. This survey aims to address this need by providing a thorough review of BioLMs, focusing on their evolution, classification, and distinguishing features, alongside a detailed examination of training methodologies, datasets, and evaluation frameworks. We explore the wide‐ranging applications of BioLMs in critical areas such as disease diagnosis, drug discovery, and vaccine development, highlighting their impact and transformative potential in bioinformatics. We identify key challenges and limitations inherent in BioLMs, including data privacy and security concerns, interpretability issues, biases in training data and model outputs, and domain adaptation complexities. Finally, we highlight emerging trends and future directions, offering valuable insights to guide researchers and clinicians toward advancing BioLMs for increasingly sophisticated biological and clinical applications.

Pontocerebellar hypoplasia (PCH) is a group of rare neurodevelopmental disorders with limited diagnostic and therapeutic options. Mutations in WDR11, a subunit of the FAM91A1 complex, have been found in patients with PCH-like symptoms; however, definitive evidence that the mutations are causal is still lacking. Here, we show that depletion of FAM91A1 results in developmental defects in zebrafish similar to that of TBC1D23, an established PCH gene. FAM91A1 and TBC1D23 directly interact with each other and cooperate to regulate endosome-to-Golgi trafficking of KIAA0319L, a protein known to regulate axonal growth. Crystal structure of the FAM91A1–TBC1D23 complex reveals that TBC1D23 binds to a conserved surface on FAM91A1 by assuming a Z-shaped conformation. More importantly, the interaction between FAM91A1 and TBC1D23 can be used to predict the risk of certain TBC1D23-associated mutations to PCH. Collectively, our study provides a molecular basis for the interaction between TBC1D23 and FAM91A1 and suggests that disrupted endosomal trafficking underlies multiple PCH subtypes.

The Health Level 7 (HL7) organization introduced the Information Sensitivity Policy Value Set with 45 sensitive data categories to facilitate the implementation of granular electronic consent technology. The goal is to allow patients to have control over the sharing of their sensitive medical records. This study represents the first attempt to explore physicians’ viewpoints on these categories. Twelve physicians participated in a survey, leading to revisions in 21 HL7 categories. They later classified 600 clinical data items through a second survey using the updated categories. Participants’ perspectives were documented, and data analysis included descriptive measures and heat maps. In the first survey, six participants suggested adding 19 new categories (e.g., personality disorder), and modifying 25 category definitions. Two new categories and sixteen revised category definitions were incorporated to support more patient-friendly content and inclusive language. Fifteen new category recommendations were addressed through a revision of category definitions (e.g., personality disorder described as a behavioral health condition). In the second survey, data categorizations led to recommendations for more categories from ten participants. Future revisions of the HL7 categories should incorporate physicians’ viewpoints, validate the categories using patient data or/and include patients’ perspectives, and develop patient-centric category specifications.

Significance The sorting nexin proteins are critical for endosomal trafficking and signaling. SNX17 and SNX27 are two SNX proteins displaying high similarity in domain structure, including a common FERM domain; however, they mediate distinct endocytic recycling pathways. We now solve this mystery by demonstrating that the FERM domains of SNX17 and SNX27 have different functions. Whereas the FERM domain of SNX17 recognizes the NPxY/NxxY motif, the same domain in SNX27 binds to a novel “DLF” motif within the N termini of SNX1/2. The interaction between SNX27 and SNX1/2 not only helps efficient retrieval of multiple cargoes but also promotes endosomal recruitment of SNX27. We further demonstrate that the SNX27–SNX1/2 interaction is crucial to neuronal growth and brain development in zebrafish.