PubMed Health⌕ Search

Biomedical subjects

Xu Liao

Publications and source records attributed to Xu Liao.

3 recordsLinked to original sources
PubMed2026

CoxFormer enables spatial omics inference with multimodal generative modeling.

Gene co-expression maps transcriptome-wide gene-gene relationships, yet high-quality estimates cover less than half the genome. Meanwhile, spatial omics either profiles restricted in situ panels or lacks cellular resolution. Extending co-expression transcriptome-wide could overcome these limitations by inferring unassayed gene expression at subcellular resolution. Here we show that CoxFormer integrates literature-derived gene knowledge with co-expression networks from bulk tissues and large-scale single-cell atlases to learn 512-dimensional representations for 32,016 human genes. These embeddings capture functional gene relationships and serve as a generative prior for spatial inference across platforms and modalities. Without requiring a matched single-cell RNA-sequencing reference, CoxFormer supports four applications beyond measured genes: histology-based expression imputation, gene activity prediction from chromatin accessibility, subcellular super-resolution inference, and pathological region detection. Together, CoxFormer extends gene embedding from gene- and cell-level tasks to whole-transcriptome spatial inference, providing a unified framework for biological analysis beyond the limited gene coverage of current spatial omics technologies.

Humans↗
PubMed2025

Benchmarking large language models for genomic knowledge with GeneTuring.

Large language models (LLMs) show promise in biomedical research, but their effectiveness for genomic inquiry remains unclear. We developed GeneTuring, a benchmark consisting of 16 genomics tasks with 1,600 curated questions, and manually evaluated 48,000 answers from ten LLM configurations, including GPT-4o (via API, ChatGPT with web access, and a custom GPT setup), GPT-3.5, Claude 3.5, Gemini Advanced, GeneGPT (both slim and full), BioGPT, and BioMedLM. A custom GPT-4o configuration integrated with NCBI APIs, developed in this study as SeqSnap, achieved the best overall performance. GPT-4o with web access and GeneGPT demonstrated complementary strengths. Our findings highlight both the promise and current limitations of LLMs in genomics, and emphasize the value of combining LLMs with domain-specific tools for robust genomic intelligence. GeneTuring offers a key resource for benchmarking and improving LLMs in biomedical research.

Benchmark↗
PubMed2004

[Tracheal intubation under general anesthesia in patients with difficult laryngoscopy].

OBJECTIVE: To evaluate the safety and efficacy of oral endotracheal intubation in the patients with difficult laryngoscopy undergoing general anesthesia. METHODS: A total of 1 683 patients with difficult laryngoscopy, aged 1.5-67 yr, and scheduled for the elective plastic surgery were observed in this study from 1989-1997. All these patients were at American Society of Anesthesiologist physical status I. According to the preoperative predictive results for difficult laryngoscopy, we classified these patients into two groups: Group I included 1 375 patients, whose epiglottis could be viewed (laryngoscopic view grades II and III); and Group II, included 308 patients, whose epiglottis could not be viewed (laryngoscopic view grade IV). For group I, anesthesia was induced with thiopentone 4-5 mg/kg and succinylcholine 1 mg/kg; Laryngoscopy was carried out using modified Macintosh method. For Group II, anesthesia was induced with a total intravenous anesthesia or inhaled anesthesia; anesthetic depth was required to effectively inhibit laryngeal reflexes with reservation of spontaneous breathing. Tracheal intubation was performed by fiberoptic stylet laryngoscope (FOSL). During anesthesia induction and tracheal intubation procedures, electrocardiogram, arterial pressure, heart rate and pulse oxygen saturation (SpO2) were continuously monitored. Complications of intubation (arrhythmia, and so on) were observed and recorded. Immediately after laryngoscopy and successful intubation, patients were examined for any traumatic injuries at teeth, lips, tongue, and oropharyngeal tissues. RESULTS: In group I, tracheal intubation was accomplished by the first attempt in 1 279 cases (93.0%) and the intubation time was less than 3 min in 1 304 cases (94.8%). In group II, tracheal intubation was accomplished by the first attempt in 114 patients (37.0%), and 123 patients (39.9%) had the intubation time of less than 3 min. Tracheal intubation was successful by the second or third attempt in 96 patients of group I and 156 patients of group II, respectively. Thirty-eight patients required four or more attempts, which only occurred in group II. Of all the complications of tracheal intubation, the traumatic complications were most common. The incidences of traumatic complications in the patients with laryngoscopic view grade II, III (group I ) and IV (group II) were 0.7%, 3.9% and 14.3%, respectively. Other complications such as respiratory depression were only seen in group II. A pooled incidence of the intubation complications was 6.7% (113/1 683). CONCLUSION: An anesthesiologist who is skillful in difficult airway management may safely manage the airway in the patients with difficult laryngoscopy under general anesthesia.

Adolescent↗