Lila Sciences

Endoscopic full-thickness resection (EFTR) has emerged as a minimally invasive treatment for small gastrointestinal subepithelial neoplasms (SETs). In particular, the development of novel closure methods such as clip-line and various endoscopic suturing techniques has enabled the secure closure of any resultant defects. During the Advanced Therapeutic Endoscopy session at the Endoscopic Forum Japan 2025, we summarized the current status of EFTR for SETs and discussed future clinical applications for epithelial neoplasms from oncological and technical perspectives. Although predominantly documented in gastric subepithelial neoplasms, EFTR is also used to treat select esophageal, duodenal, and colorectal SETs. While extensively evaluated for gastric gastrointestinal stromal tumors (GISTs), relatively low R0 resection rates have limited its broader clinical adoption; however, the novel no-touch EFTR technique could overcome this challenge and facilitate the adoption of this procedure for small gastric GISTs. Applying EFTR to epithelial tumors requires strict assessment of oncological clearance, given the risks of lymph node metastasis and peritoneal seeding following full-thickness breaches. At the end of the session, we voted on the appropriateness of EFTR for epithelial neoplasms across different organs, balancing oncological risks against technical feasibility. All nine participants (100%) supported EFTR for rectal lesions, and eight (89%) supported its use for gastric lesions. Future investigations and technical innovations are required to expand the clinical indications for EFTR in epithelial neoplasms.

To advance the computational simulation of cellular life, we propose a virtual yeast, an artificial intelligence (AI)-driven agent that models eukaryotic cellular behaviours by integrating multimodal biological data, mechanistic reasoning and active experimentation using Saccharomyces cerevisiae as a genetically tractable and data-rich model system. Cellular complexity is decomposed into eight function-centred modules, spanning genetic, metabolic and structural systems, each realized as a domain-specific AI tool coordinated through a large language model-based orchestration layer. Built on three data pillars, namely, mechanistic knowledge, subcellular architecture and dynamic states, the system integrates representation learning and generative modelling within a closed-loop learning pipeline that autonomously designs and executes experiments. The virtual yeast serves as both a conceptual and an operational platform to optimize biosynthetic pathways, support the generation and prioritization of hypotheses across diverse cellular processes, and accelerate target discovery. By coupling biological realism with autonomous AI reasoning, the virtual yeast establishes a generalizable blueprint for constructing virtual eukaryotic cells and advancing synthetic biology.

Enhancers are key regulators of mammalian gene expression, yet how they interact with promoters in space (contact vs. action-at-a-distance) and in time (transient vs. stable) remains poorly understood. Recent studies suggest that enhancers can activate promoters across distances exceeding 200 nanometers, challenging classical contact models, but limited spatiotemporal resolution has obscured the mechanistic details of enhancer-promoter (E-P) interactions and their link to transcription. Here, we engineered a synthetic biology platform optimized for the simultaneous visualization of E-P 3D distance and nascent transcription using super-resolution live-cell imaging. By applying five complementary approaches integrating imaging, 3D genomics, and gene expression data across cell lines, we estimate that transcriptional activation is mediated by ~25-42 nanometer contacts on the seconds timescale. Our results support a transient contact mechanism for E-P-mediated gene activation.

Microbial cell factories represent sustainable platforms for the production of fuels, chemicals, and therapeutics, but their development is limited by challenges in pathway discovery, enzyme optimization, and metabolic regulation. Recent advances in artificial intelligence and machine learning are reshaping this field by enabling predictive pathway design, enhanced protein engineering, and dynamic network regulation. Emerging strategies such as graph neural networks, generative models, and reinforcement learning (RL) now allow systematic exploration of vast design spaces with enhanced accuracy and scalability. This review highlights recent advancements in microbial engineering. It discusses how AI-driven frameworks are advancing the field from experience-guided and rule-based engineering toward data-driven, model-assisted, and increasingly autonomous workflows. These changes lay the foundation for next-generation biomanufacturing.

Glycosylation is a critical structural modification that shapes the pharmacological properties of bioactive ingredients from Traditional Chinese Medicine (TCM), and UDP-glycosyltransferases (UGTs) are the core rate-limiting biocatalysts mediating this process. Traditional plant extraction methods are constrained by resource scarcity, long growth cycles, low target content and high environmental costs, which cannot meet the large-scale industrial demand for high-value medicinal glycosides. This review systematically outlines the latest global advances in medicinal plant UGT research, covering family classification and physiological functions, multi-omics and AI-assisted gene mining, molecular basis of substrate recognition and catalytic specificity, protein engineering for performance optimization, and the construction of full-spectrum biomanufacturing systems including in vitro multi-enzyme cascades, microbial cell factories and plant suspension cell cultures. We further discuss the core challenges of industrial scale-up, regulatory compliance and clinical translation, as well as the significant economic and technical advantages of synthetic biology-based UGT biomanufacturing platforms. This work provides a complete technical framework for the engineering application of medicinal plant UGTs, to support the green and scalable production of rare natural therapeutic glycosides.

Conventional tumor marker detection technologies are confined to single-index sequential testing, with drawbacks like prolonged turnaround times, high sample consumption, and heavy reliance on large precision instruments. They fail to meet the needs of clinical scenarios such as point-of-care testing (POCT), primary medical institution screening, and emergency rapid assessment. This study aims to develop a multi-index simultaneous quantitative detection system, providing an efficient, convenient, and reliable technical solution for early accurate diagnosis and whole-course dynamic monitoring of tumors. Leveraging the core advantages of micro fluidic chips-miniaturization, integration, and low reagent consumption-a detection platform was designed to simultaneously quantify six tumor-associated markers: matrix metalloproteinase 9 (MMP-9), vascular endothelial growth factor A (VEGF-A), soluble neural cadherin (sN-cadherin), osteoprotegerin (OPG), lysyl oxidase (LOX), and angiopoietin 2 (ANG-2). Methodological characterization included linear range verification, limit of detection (LOD) determination, precision evaluation, and specificity tests (cross-reactivity, matrix interference, and background interference). Consistency with clinical gold-standard methods was assessed via correlation analysis, Kappa test, and Bland-Altman analysis. Diagnostic efficacy was evaluated using ROC curve analysis, and detection timeliness was improved by optimizing the "two-reaction and two-washing" core process. All six markers exhibited excellent analytical performance: the coefficients of determination (R²) of their dose-response curves ranged from 0.9968 to 0.9993, with linear ranges of 0.012-16,000 pg/mL (VEGF-A, OPG, ANG-2) and 0.016-9,600 ng/mL (MMP-9, sN-cadherin, LOX), and limits of detection (LODs) of 0.012-0.019 pg/mL (or equivalent ng/mL units). Precision was outstanding: intra-batch relative standard deviations (RSDs) were 2.51%-5.12% for low-concentration samples and 0.88%-4.14% for high-concentration samples, while inter-batch RSDs, chip repeatability RSDs, and storage stability RSDs were all ≤ 6.76%, meeting the clinical threshold standard of ≤10%. Specificity verification showed that both cross-reactivity rates and interference rates were significant non-specific binding or matrix interference observed. Compared with the gold standard, the coefficient of determination (R²) was > 0.95, the Kappa coefficient was 0.8-1.0 (excellent agreement), and over 90% of sample deviations fell within the 95% confidence interval (CI). The area under the receiver operating characteristic curve (AUC) ranged from 0.9546 to 0.9882, with both detection sensitivity and specificity reaching 93%-98%. The detection system shortened the total detection time to 24 minutes, required only microliter-scale sample consumption, and was 5-8 times faster than enzyme-linked immunosorbent assay (ELISA)/chemiluminescence immunoassay. The microfluidic chip system integrates high sensitivity, precision, specificity, rapidity, and miniaturization, achieving high equivalence with traditional methods. It breaks through conventional limitations, meeting clinical needs for early screening, dynamic monitoring, and large-scale surveillance, with significant clinical transformation and industrialization prospects. Future optimization will involve multi-center validation, panel expansion, and AI integration to support personalized tumor care.

Endoscopic local triamcinolone injection is a standard therapy for esophageal stricture (ES) prevention after extensive endoscopic submucosal dissection (ESD); however, oral prednisolone administration is a promising alternative. This trial aimed to investigate the superiority of oral prednisolone administration to local triamcinolone injection for ES prevention after extensive ESD. This phase 3, open-label, multicenter, randomized controlled trial enrolled patients with extensive superficial esophageal squamous cell carcinoma. The primary endpoint was stricture-free survival (SFS). Secondary endpoints were the number of endoscopic balloon dilations (EBDs) for 12 weeks after ESD, dysphagia score ≤ 1 at 12 weeks after ESD, and adverse events (AEs). A total of 281 patients from 45 institutions were enrolled. The 12-week SFS rates were 88.5% (95% confidence interval [CI]: 81.6-92.9) and 94.8% (95% CI: 89.4-97.5) in the triamcinolone and prednisolone arms, respectively (hazard ratio 0.672, 90% CI [0.361-1.250], p = 0.14). Thirty-eight and 26 EBDs were performed for 14 and 10 patients, respectively, in the triamcinolone and prednisolone arms within 12 weeks after ESD. In the triamcinolone and prednisolone arms, 73.8% (95% CI: 65.7-80.8) and 80.7% (95% CI: 73.2-86.9) of patients, respectively, had dysphagia scores ≤ 1 at 12 weeks after ESD (p = 0.20). Grade 3/4 AEs occurred in 12 (8.7%) and 9 patients (6.7%) in the triamcinolone and prednisolone arms, respectively. While oral prednisolone administration and local triamcinolone injection demonstrated comparable outcomes, oral prednisolone did not demonstrate superiority for ES prevention after extensive esophageal ESD.