¸£Àû×ÅƬ

This study demonstrates the feasibility of using large language models to automatically structure unstructured histology and radiology reports in IBD at scale. Evaluating over 32,000 records, the authors show that LLMs can reliably extract clinically meaningful, standardised data, outperforming manual curation in key tasks. The work supports scalable, privacy-compliant AI solutions for FAIR data generation in IBD research and clinical translation.

 Using longitudinal real-world data from over 1,500 patients with IBD, this study shows that routinely collected biomarkers at IBD diagnosis—particularly ALT and ALP—can identify individuals at increased risk of IBD-associated liver disease. Distinct biomarker trajectories precede diagnosis, supporting earlier investigation and closer monitoring to reduce diagnostic delay and improve outcomes.

 This benchmarking study compares the performance of widely used whole-exome sequencing kits, demonstrating that capture efficiency and target design substantially influence coverage of clinically relevant coding regions. Twist exome kits showed particularly strong performance despite smaller target sizes. The findings provide practical guidance for researchers selecting exome platforms for clinical and translational genomics.

Analysing two large IBD cohorts, this study identifies an altered burden of functionally impactful variants in ERAP1 and ERAP2 genes, particularly in ulcerative colitis and UC with autoimmune comorbidity. Variant- and gene-level analyses support a role for antigen-processing pathways in IBD susceptibility, linking evolutionary selection with modern autoimmune disease risk.