SEARCH · PubMed Health

Results for “Benchmark”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 55 records · Page 3Linked to original sources

PubMed2026

Benchmarking urinary cell transcriptomes for noninvasive differentiation of BK polyomavirus-associated nephropathy from T cell-mediated rejection.

BK polyomavirus-associated nephropathy (BKVN) adversely impacts kidney allograft survival and often mimics acute T cell-mediated rejection (TCMR), confounding diagnosis and management. To address this conundrum, we performed unbiased RNA sequencing of urinary cells matched to biopsies classified as BKVN with intragraft inflammation (BKVN-P), BKVN without inflammation (BKVN-N), TCMR, or no rejection (NR). BKVN-N displayed dominant host DNA replication, cell cycle, and repair programs, while BKVN-P samples exhibited expansive innate immune activation, antigen presentation, chemokine upregulation, and epithelial injury. Both BKVN subtypes shared signatures of T cell exhaustion and mature and tolerogenic dendritic cell activation but differed in immune orientation - Th1 predominance in BKVN-N versus Treg and CD8 enrichment in BKVN-P. Compared with TCMR samples, BKVN-P lacked robust TCR/CD28 signaling and was enriched for viral and innate modules; BKVN-N lacked alloimmune activation. B cell exhaustion characterized BKVN-N, while BKVN-P displayed robust B cell activation with metabolic downregulation. A ratiometric urinary cell biomarker, CXCL10 mRNA/CD3E mRNA, distinguished both BKVN subtypes from TCMR with diagnostic accuracy, replicated by quantitative reverse transcription PCR for clinical translation, and confirmed in an independent cohort. These findings demonstrate the utility of urinary cell transcriptomics for resolving viral injury from alloimmunity, enabling precision diagnostics and targeted immunomodulation in kidney transplantation.

Humans↗

PubMed2025

Refined and benchmarked homemade media for cost-effective, weekend-free human pluripotent stem cell culture.

BACKGROUND: Cost-effective, practical, and reproducible culture of human pluripotent stem cells (hPSCs) is required for basic and translational research. Basal 8 (B8) has emerged as a cost-effective solution for weekend-free and chemically-defined hPSC culture. However, the requirement to home-produce some recombinant growth factors for B8 can hinder access and reproducibility. Moreover, we found the published B8 formulation suboptimal in widely-used normoxic hPSC culture. Lastly, the performance of B8 in functional applications such as genome editing or organoid differentiation required systematic evaluation. METHODS: We formulated B8 with commercially available, growth factors and adjusted its composition to support normoxic culture of WTC11 human induced pluripotent stem cell line. We compared this formulation (B8+) with commercial Essential 8 (cE8) and a home-made, weekend-free E8 formulation (hE8). We measured pluripotency marker expression and cell cycle by flow cytometry, and investigated the transcriptional profiles by bulk and single-cell RNA sequencing. We further assessed genomic stability, genome editing efficiency, single-cell cloning, and differentiation in both monolayer and organoids. Finally, we validated key findings using male (H1) and female (H9) human embryonic stem cells. RESULTS: hE8 performed comparably to cE8 across most functional assays and cell lines. In contrast, cells in B8+ displayed higher NANOG expression and improved genome editing efficiency. At the same time, B8+ led to gene expression changes indicative of marked lineage priming, reflected in altered morphology and differential response to some differentiation protocols. Both weekend-free media resulted in a modest transcriptional shift towards a less metabolically active state, consistent with intermittent media starvation. CONCLUSIONS: Homemade weekend-free media can provide a cost-effective alternative to commercial formulations. hE8, integrating some features of B8 while resembling cE8, emerges as a robust and practical option with limited compromises. B8+, though advantageous in some contexts, warrants caution due to lineage priming effects that may impact differentiation outcomes.

hiPSC; pluripotency; culture media; thermostable F↗

PubMed2022

Improving risk indexes for Alzheimer's disease and related dementias for use in midlife.

Knowledge of a person's risk for Alzheimer's disease and related dementias (ADRDs) is required to triage candidates for preventive interventions, surveillance, and treatment trials. ADRD risk indexes exist for this purpose, but each includes only a subset of known risk factors. Information missing from published indexes could improve risk prediction. In the Dunedin Study of a population-representative New Zealand-based birth cohort followed to midlife (N = 938, 49.5% female), we compared associations of four leading risk indexes with midlife antecedents of ADRD against a novel benchmark index comprised of nearly all known ADRD risk factors, the Dunedin ADRD Risk Benchmark (DunedinARB). Existing indexes included the Cardiovascular Risk Factors, Aging, and Dementia index (CAIDE), LIfestyle for BRAin health index (LIBRA), Australian National University Alzheimer's Disease Risk Index (ANU-ADRI), and risks selected by the Lancet Commission on Dementia. The Dunedin benchmark was comprised of 48 separate indicators of risk organized into 10 conceptually distinct risk domains. Midlife antecedents of ADRD treated as outcome measures included age-45 measures of brain structural integrity [magnetic resonance imaging-assessed: (i) machine-learning-algorithm-estimated brain age, (ii) log-transformed volume of white matter hyperintensities, and (iii) mean grey matter volume of the hippocampus] and measures of brain functional integrity [(i) objective cognitive function assessed via the Wechsler Adult Intelligence Scale-IV, (ii) subjective problems in everyday cognitive function, and (iii) objective cognitive decline measured as residualized change in cognitive scores from childhood to midlife on matched Weschler Intelligence scales]. All indexes were quantitatively distributed and proved informative about midlife antecedents of ADRD, including algorithm-estimated brain age (β's from 0.16 to 0.22), white matter hyperintensities volume (β's from 0.16 to 0.19), hippocampal volume (β's from -0.08 to -0.11), tested cognitive deficits (β's from -0.36 to -0.49), everyday cognitive problems (β's from 0.14 to 0.38), and longitudinal cognitive decline (β's from -0.18 to -0.26). Existing indexes compared favourably to the comprehensive benchmark in their association with the brain structural integrity measures but were outperformed in their association with the functional integrity measures, particularly subjective cognitive problems and tested cognitive decline. Results indicated that existing indexes could be improved with targeted additions, particularly of measures assessing socioeconomic status, physical and sensory function, epigenetic aging, and subjective overall health. Existing premorbid ADRD risk indexes perform well in identifying linear gradients of risk among members of the general population at midlife, even when they include only a small subset of potential risk factors. They could be improved, however, with targeted additions to more holistically capture the different facets of risk for this multiply determined, age-related disease.

Alzheimer’s disease↗

PubMed2026

Precision targeting of teacher burnout using network-informed ecological momentary interventions.

Teacher well-being affects classroom functioning and workforce stability, yet generic digital programs rarely use person-specific affect dynamics to select support. This cluster-randomised trial evaluated whether micro-interventions selected from high expected influence (EI) nodes in teachers' contemporaneous affect networks produced larger changes in burnout-related EI and everyday happiness than content-matched random allocation. The objectives were to estimate allocation effects on changes in estimated network summaries and happiness, evaluate network change as a statistical mediator, examine personality moderation, and benchmark simpler allocation rules. A two-arm cluster randomised platform trial was conducted in 84 public schools across four urban districts in H Province. After a 14 day baseline of ecological momentary assessment (EMA), person specific partial correlation networks were estimated for happiness, exhaustion, detachment, efficacy and rumination. An optimisation engine prioritised three brief micro-intervention types per teacher according to baseline EI, while the active control received the same library without network information. EMA continued for 8 weeks; Bayesian multilevel models, permutation-based mediation, and benchmarking analyses were applied. EI-based targeting produced larger reductions in the composite EI-change index than active control (mean difference 0.11, 95% credible interval 0.08 to 0.14) and higher week 7 EMA happiness (4.4 points on a 0 to 100 scale, 95% credible interval 2.7 to 6.0), with a positive arm by week slope difference of 0.62 points per week (95% credible interval 0.39 to 0.85). Model-based mediation estimates were consistent with approximately one half of the happiness difference being statistically associated with change in the composite EI-change index (average conditional mediation estimate 3.5 points, 95% credible interval 2.0 to 5.2). Benchmarking showed smaller gains under severity, threshold, or group-level centrality rules. Effects were stronger among teachers higher in conscientiousness. The findings indicate that integrating EMA, network modelling, and EI-driven optimisation yields measurable gains beyond content-matched exposure, providing a proof of concept for district-scale precision mental health that requires prospective implementation testing. Replication in additional regions, expanded node sets, and longer follow up are warranted to assess durability and generalisability.

Female↗

PubMed2026

Current Diagnostic Pathways for Rheumatoid Arthritis-Associated Interstitial Lung Disease Result in Substantial Underdiagnosis and Excess Mortality: A Multicenter Norwegian Quality Assurance Audit.

OBJECTIVE: Recent guidelines suggest risk-stratified screening for rheumatoid arthritis-associated interstitial lung disease (RA-ILD). However, the diagnostic gap between current routine care and this screening approach remains unquantified. We assessed currently detected RA-ILD in Norway, benchmarking findings against recent screening-based estimates of the true disease burden. METHODS: This 10-year quality assurance audit across six centers covered 43% of the Norwegian population. RA-ILD cases identified via ICD-10 codes were confirmed by manual chart review. Prevalence was calculated relative to a registry-derived total RA background population and benchmarked against a 10% expected target derived from recent prospective studies. Mortality was compared to a 3:1 frequency-matched RA control group using Cox proportional hazards regression. RESULTS: Among 17,305 RA patients, 188 (1.1%) had verified ILD; when benchmarked against an expected 10% prevalence, this indicates an 89% diagnostic gap in routine clinical care. Mean age at ILD detection was 67.5 years. Most cases (93.6%) possessed ≥2 established risk factors for RA-ILD: 93.6% were seropositive, 76.1% had smoking histories, while RA onset age ≥60 and persistently increased inflammatory laboratory markers were present in over half of patients. RA-ILD was associated with significantly increased mortality; 66 (4.1/100 person-years) deaths occurred in the RA-ILD group vs. 120 (2.3/100 person-years) among RA controls (HR 1.77; 95% CI: 1.31-2.39, p<0.001). CONCLUSION: When comparing to prevalence expectations, current routine care may leave a substantial proportion of cases undetected, primarily capturing a high-risk phenotype with excess mortality. Systematic, risk-stratified screening is needed to bridge this diagnostic gap, aiming to enable earlier intervention.

Interstitial lung disease↗

PubMed2024

metaExpertPro: A Computational Workflow for Metaproteomics Spectral Library Construction and Data-Independent Acquisition Mass Spectrometry Data Analysis.

Analysis of large-scale data-independent acquisition mass spectrometry metaproteomics data remains a computational challenge. Here, we present a computational pipeline called metaExpertPro for metaproteomics data analysis. This pipeline encompasses spectral library generation using data-dependent acquisition MS, protein identification and quantification using data-independent acquisition mass spectrometry, functional and taxonomic annotation, as well as quantitative matrix generation for both microbiota and hosts. By integrating FragPipe and DIA-NN, metaExpertPro offers compatibility with both Orbitrap and timsTOF MS instruments. To evaluate the depth and accuracy of identification and quantification, we conducted extensive assessments using human fecal samples and benchmark tests. Performance tests conducted on human fecal samples indicated that metaExpertPro quantified an average of 45,000 peptides in a 60-min diaPASEF injection. Notably, metaExpertPro outperformed three existing software tools by characterizing a higher number of peptides and proteins. Importantly, metaExpertPro maintained a low factual false discovery rate of approximately 5% for protein groups across four benchmark tests. Applying a filter of five peptides per genus, metaExpertPro achieved relatively high accuracy (F-score = 0.67-0.90) in genus diversity and showed a high correlation (rSpearman = 0.73-0.82) between the measured and true genus relative abundance in benchmark tests. Additionally, the quantitative results at the protein, taxonomy, and function levels exhibited high reproducibility and consistency across the commonly adopted public human gut microbial protein databases IGC and UHGP. In a metaproteomic analysis of dyslipidemia patients, metaExpertPro revealed characteristic alterations in microbial functions and potential interactions between the microbiota and the host.

Proteomics↗

PubMed2026

Prediction of human missense variant effects from functional evidence.

Prediction of missense variant effects remains a critical bottleneck in both research and diagnostic genetics. Current predictors typically rely on clinical outcomes or population patterns rather than direct measures of functional impact, leading to limited generalizability and data circularity. Here we present FuncVEP, a family of variant effect predictors trained on diverse functional data to predict the functional impact of missense variants. FuncVEP generalizes across datasets and outperforms 48 existing predictors across a wide range of benchmarks, improving accuracy from 78.8% to 84.6% on functional benchmarks and from 90.1% to 92.4% on clinical benchmarks. From a discovery perspective, we identified 210 new gene-phenotype associations involving 494 genes linked to inborn errors of immunity in the UK Biobank and the Mount Sinai Million Health Discoveries Program. FuncVEP substantially improved the discovery rate relative to state-of-the-art predictors. Overall, FuncVEP provides a robust, scalable solution for variant interpretation, advancing both diagnostic precision and gene discovery.

Humans↗

PubMed2026

Agentomics: an agentic system that autonomously develops novel state-of-the-art solutions for biomedical machine learning tasks.

MOTIVATION: Extracting knowledge from biomedical data is crucial for advancing our understanding of biological systems and developing novel therapeutics. The quantity, quality, and resolution of biomedical data constantly evolves, requiring the automation of biomedical machine learning (ML). Existing Automated ML tools lack flexibility, while large language models (LLMs) struggle to consistently deliver reproducible machine learning codebases, and existing LLM Agent-powered solutions lag behind human-engineered ML models. RESULTS: Here, we introduce Agentomics, an autonomous LLM-powered agentic system for end-to-end ML experimentation. Given a biomedical dataset, Agentomics implements various ML modeling strategies, and produces a ready-to-use ML model. Agentomics introduces strict validation checkpoints for standard ML development steps, allowing gradual development on top of working code with defined interfaces and validated artifacts. Further, it offers native support for biomedical foundation models that can be leveraged during experimentation. The generic nature of Agentomics allows the user to create ML solutions for a large variety of datasets and use various LLMs. We evaluate Agentomics across 20 datasets from the domains of Protein Engineering, Drug Discovery, and Regulatory Genomics. When benchmarked against other agentic systems, Agentomics outperformed them in all tested domains. When benchmarked against human expert solutions, Agentomics generated novel state-of-the-art models for 11/20 established benchmark datasets. AVAILABILITY AND IMPLEMENTATION: Agentomics is implemented in Python. Source code and documentation are freely available at: https://github.com/BioGeMT/Agentomics-ML.

Machine Learning↗

PubMed2026

Network methods for diagonal integration of unpaired single-cell multiomics data: a review.

MOTIVATION: Advances in single-cell sequencing have enabled multiomics profiling at unprecedented resolution; however, mass spectrometry-based single-cell proteomics (scMS) remains inherently destructive, precluding simultaneous transcriptomic capture. Unlike antibody-based methods such as CITE-seq, which permit paired profiling but are restricted to targeted protein panels, scMS provides unbiased, genome-scale coverage of the intracellular proteome yet necessitates post hoc integration of unpaired datasets. This diagonal integration challenge, where transcriptomes and proteomes are measured in separate cells lacking shared anchors, remains underserved by existing reviews, which focus predominantly on vertical integration strategies enabled by non-destructive assays. RESULTS: We survey the complete computational pipeline for constructing mechanistic proteogenomic networks from unpaired single-cell data, covering: (i) unimodal network inference such as knowledge-based approaches, probabilistic graphical models, temporal directionality inference, and generative and foundation model strategies that establish the transcriptomic scaffold; (ii) cross-modal integration architectures such as network propagation, graph neural networks (scMRDR, scmFormer, scCotag), and consensus frameworks designed explicitly for the unpaired proteomics setting; and (iii) benchmarking paradigms spanning network reconstruction (BEELINE, GRETA, CausalBench) and multi-task integration evaluation (scMultiBench, SCMMIB), with guidance on metric selection under network sparsity and class imbalance. We identify three principal axes of future development: generative proteomic translation from transcriptomic precursors, inductive prior embedding in next-generation architectures, and perturbation-based causal benchmarking. AVAILABILITY AND IMPLEMENTATION: This is a review article; no novel software is distributed. A curated benchmark resource table, methods starter guide, and per-method bottleneck annotations are provided in the Supplementary Material.

Multiomics↗

PubMed2026

GUANinE v1.1 reveals complementarity of supervised and genomic language models.

There has been much debate about the benefits of supervised versus unsupervised learning on genomes. Determining which is better in what contexts requires developing comprehensive benchmarks spanning functional and evolutionary tasks. Importantly, such benchmarks need large sample sizes to enable well-powered ranking of models. Having developed and applied such a benchmark here (GUANinE v1.1), we conclusively demonstrate each paradigm offers key advantages and outperforms on certain tasks. In accordance with training, supervised sequence-to-function models exhibit strong performance when annotating functional states characterized by chromatin accessibility or histone marks, while self-supervised language models outperform on evolutionary conservation. Our hundreds of new evaluations in this v1.1 expansion provide evidence for a tradeoff between input context size and model parameter count for a fixed compute budget, which we depict with new metrics such as kiloparameters/base pair. We also construct two new large-scale variant interpretation tasks in v1.1: cadd-snv measuring deleteriousness, and clinvar-snv measuring clinical pathogenicity. We find that conservation scores, and by extension, genomic language models, predict deleteriousness well, but successfully translating deleteriousness predictions to pathogenicity remains challenging. GUANinE v1.1 newly evaluates dozens of pretrained genomic models, and we conclude that moderate-context hybrid or post-trained language models may define the next era of machine learning in genomics.

Genomics↗

PubMed2026

AI-Driven Multi-Omics Integration of Synthetic Colon Adenocarcinoma for Cluster-Guided PROTAC Candidate Design Targeting KRASG12D.

Colorectal cancer is a leading cause of cancer death, yet its molecular heterogeneity remains poorly translated into individualized treatment. We present a reproducible artificial intelligence (AI) framework that integrates multi-omics benchmarking, sample-level drug prioritization, E3 ubiquitin ligase selection, and shape-anchored Proteolysis Targeting Chimera (PROTAC) design for KRASG12D in colon adenocarcinoma (COAD). A controlled synthetic benchmark comprising 425 tumor and 41 simulated normal profiles, parameterized to match The Cancer Genome Atlas (TCGA) distributions, was used for pipeline verification. Among sixteen methods, the Balanced Latent Integration with Stability Selection (BLISS) model achieved the highest silhouette width (0.86) and competitive agreement (Adjusted Rand Index, ARI, 0.90). The pipeline was validated on real data: a TCGA COAD cohort (186 tumors) with independent Consensus Molecular Subtype (CMS) labels and a CPTAC cohort (104 tumors). Integration modestly recovered CMS (ARI 0.28), and stage, not molecular cluster, drove survival (log-rank p = 0.005 versus 0.81). Sample-level prioritization differed from cluster-level ranking in 82.6% of profiles, below chance (p < 0.0001), without indicating efficacy. Candidate NOVEL00489 showed a good MM-GBSA estimate, matching the reference ASP3082. Compounds are computational candidates requiring experimental validation. This establishes a transparent benchmark for in silico degrader generation in precision oncology.

Humans↗

PubMed2025

In silico generation of synthetic cancer genomes using generative AI.

Understanding how genomic alterations drive cancer is key to advancing precision oncology. To detect these alterations, accurate algorithms are used; however, due to privacy concerns, few deeply sequenced cancer genomes can be shared, limiting benchmarking and representing a major obstacle to the improvement of analytic tools. To address this, we developed OncoGAN, a generative AI model combining adversarial networks and variational autoencoders to create realistic synthetic cancer genomes. Trained on large-scale genomic datasets, OncoGAN accurately reproduces somatic mutations, copy number alterations, and structural variants across cancer types while preserving donors' privacy. The synthetic genomes reflect tumor-specific mutational signatures and positional mutation patterns. Using DeepTumour, we validated the synthetic data's fidelity, showing high concordance between generated and predicted tumors. Moreover, augmenting the training data with synthetic genomes improved DeepTumour's accuracy, underscoring OncoGAN's potential to generate shareable datasets with known ground truths for benchmarking and enhancement of cancer genome analysis tools.

Humans↗

PubMed2026

Exploring differences across pangenome-graph representations using Escherichia coli O157:H7 as a model.

Pangenome graphs are increasingly used to represent population-scale bacterial diversity, yet construction methods span fundamentally different representation paradigms whose outputs and sensitivities to assembly quality remain poorly quantified. We systematically reviewed microbial pangenome graph tools and benchmarked seven representative methods spanning gene-cluster, compacted coloured de Bruijn graph, one hybrid approach and one multiple sequence alignment method. Using a repeat-rich Escherichia coli O157:H7 dataset with complete genomes and matched short-read data, we constructed graphs from identical inputs and observed orders-of-magnitude differences in graph size and fragmentation, indicating that global topology is driven by representation strategy. Varying completeness composition revealed that assembly fragmentation is a first-order determinant of graph structure: gene-cluster graphs contracted as draft assemblies replaced complete genomes, whereas compacted coloured de Bruijn graphs expanded, with distinct degree-prevalence fingerprints across tools. In contrast, the multiple sequence alignment method could not be evaluated across fragmented inputs because it did not run reliably on draft-assembly datasets. Computational cost mirrored these shifts and depended strongly on completeness composition, including a pronounced runtime penalty for one compacted coloured de Bruijn graph implementation on all-draft inputs. Finally, analysis of Shiga toxin loci showed that pangenome-level reconciliation by gene-cluster-based tools does not reliably correct assembly artefacts at challenging multi-copy genes and that performance varies by locus. Together, these findings show that pangenome graphs are representation-dependent models of bacterial diversity, and that, in this repeat-rich O157:H7 benchmark dataset, assembly completeness is a primary determinant of their topology, scalability, and locus-level accuracy.

Escherichia coli O157↗

PubMed2026

An encyclopedia of human enhancer-gene regulatory interactions.

Identifying transcriptional enhancers and their target genes is essential for understanding gene regulation and the effect of human genetic variation on disease1-6. Here we create and evaluate a resource of more than 92 million enhancer-gene regulatory interactions across 1,458 biosamples covering 369 cell types and tissues, by integrating predictive models, chromatin states, three-dimensional contacts and large-scale genetic perturbations generated by the ENCODE Consortium7. We first create a systematic benchmarking pipeline to compare predictive models, assembling a dataset of 10,356 element-gene pairs measured in CRISPR perturbation experiments, more than 30,000 fine-mapped expression quantitative trait loci and 569 fine-mapped genome-wide association study (GWAS) variants linked to a probable causal gene. Using this framework, we develop ENCODE-rE2G, a predictive model achieving state-of-the-art performance across several prediction tasks, demonstrating that iterative perturbations and supervised machine learning can build increasingly accurate predictive models of enhancer regulation. Using ENCODE-rE2G, we build an encyclopedia of enhancer-gene regulatory interactions in the human genome, revealing global properties of enhancer networks, identifying differences in regulatory complexity across genes and improving analyses linking noncoding variants to target genes and cell types for common complex diseases. By interpreting the model, we find that beyond enhancer activity and three-dimensional enhancer-promoter contacts, additional features that guide enhancer-promoter communication include promoter class and enhancer-enhancer synergy. These genome-wide maps of enhancer-gene regulatory interactions, benchmarking software, predictive models and insights about enhancer function provide a valuable resource for future studies of gene regulation and human genetics.

Humans↗

PubMed2026

Structured robotic colorectal training in a non-tertiary NHS hospital: a 502-case consecutive cohort implementation study.

Robotic-assisted colorectal surgery has expanded rapidly across NHS practice in the UK. Structured unit-wide training pathways are essential for safe technology adoption, yet published outcome data from non-tertiary hospitals remain limited. This study describes the implementation and feasibility of a unit-wide robotic colorectal program at a high-volume non-tertiary hospital, reporting outcomes across 502 consecutive resections performed by eight consultant surgeons and presenting these in the context of nationally published benchmarks. A retrospective cohort study of 502 consecutive robotic colorectal resections performed at York Teaching Hospital between May 2022 and December 2025. Eight consultant surgeons (A-H) participated in a structured four-phase training pathway incorporating simulation training, proctored cases, complexity-based case progression, and formal credentialing. Primary outcomes were 30-day mortality, unplanned return to theatre (RTT), and anastomotic leak (AL). Anastomotic leak was calculated using only patients who underwent anastomosis as the denominator. Procedure-stratified and individual surgeon outcomes with 95% confidence intervals were reported. Risk-adjusted cumulative sum (RA-CUSUM) analysis was performed to evaluate learning curves. Outcomes are presented descriptively alongside nationally published reference data; no formal statistical comparison against national benchmarks was performed. 502 robotic colorectal resections were performed. Mean patient age was 70.0 ± 11.3 years; 58.4% were male. Median ASA grade was III. The indication was malignancy in 89.2% of cases. Length of stay was non-normally distributed and is therefore reported using median and interquartile range in the revised analysis. Key outcomes: - 30-day mortality: 1.0% (5/502; 95% CI 0.4-2.3%) - Unplanned return to theatre (RTT): 5.2% (26/502; 95% CI 3.6-7.5%) - Anastomotic leak (AL): 3.3% (15/450; 95% CI 2.0-5.5%; denominator = patients with anastomosis) - 30-day unplanned readmission: 5.0% (25/502; 95% CI 3.4-7.2%) - Conversion to open surgery: 3.6% (18/502; 95% CI 2.3-5.6%) - Lymph node yield ≥12: 91.3% of cancer resections - R0 resection rate: 95.1% of cancer resections All primary outcomes fell within or below the published reference ranges used for descriptive context. RA-CUSUM trajectories were heterogeneous: no surgeon crossed the predefined upper control limit, but several curves showed later upward movement. Accordingly, the analysis is interpreted as safety surveillance rather than evidence of uniform performance improvement. RA-CUSUM monitoring showed that no surgeon crossed the predefined upper control limit; however, heterogeneous trajectories precluded a claim of uniform performance improvement.

Humans↗

PubMed2026

Reinforcement learning-based dynamic ensemble for missense variant effect prediction and tiered prioritization of VUS.

BACKGROUND: Accurate classification of missense variants remains a challenging task despite major advances in genomics. Numerous computational models have been developed to assist in variant classification, but often require repeated integration and benchmarking efforts. Ensemble methods have been proposed to overcome the limitations of single predictors, but mostly rely on fixed, predefined weights that constrain their ability to capture interactions among predictive signals. METHODS: We present GenixRL, a dynamic ensemble framework that reformulates model fusion as a reinforcement learning optimization problem. GenixRL uses a Q-learning agent to learn a policy that dynamically weights the probabilistic outputs of complementary predictors, including BayesDel (addAF and noAF), ClinPred, and MetaRNN. Replacing static weighting with policy learning allows GenixRL to adaptively identify optimal weightings and substantially improve classification accuracy. RESULTS: In benchmark evaluation against 25 state-of-the-art predictors, GenixRL achieved an AUROC of 0.9644 on an independent ClinVar dataset. On saturation genome editing assays for BRCA1 and BRCA2, GenixRL achieved the best performance and ranked highest on 14 of 17 clinically significant genes in a zero-shot evaluation. Applied to uncertain and conflicting ClinVar variants, GenixRL enabled tiered, evidence-based prioritization of hundreds of thousands of variants as likely pathogenic or pathogenic with high confidence, supported by orthogonal population evidence from gnomAD. CONCLUSION: GenixRL advances pathogenicity prediction for missense variants and provides an adaptive ensemble that sorts variants of uncertain significance into tiered candidates for expert curation and functional validation.

Mutation, Missense↗

PubMed2026

Meta-PseU: A meta-classifier for robust prediction of RNA pseudouridine modification sites from long sequences.

BACKGROUND AND OBJECTIVES: Pseudouridine (Ψ) represents one of the most abundant and conserved RNA modifications. Ψ provides an additional hydrogen-bond donor that enhances RNA structural stability and modulates translation. It participates in diverse biological processes, including RNA-protein interactions, splicing, translational control, and stress responses. Aberrant pseudouridylation is implicated in cancer, neurodegenerative disorders, and autoimmune diseases. Despite its biological importance, experimental identification of Ψ sites remains time-consuming and costly, limiting the feasibility of transcriptome-wide profiling. Computational approaches have therefore become essential complements to experimental techniques. However, state-of-the-art machine-learning and deep-learning predictors often suffer from limited generalizability due to small training datasets. To overcome these issues, we aim at constructing new long-sequence datasets and developing a novel Ψ site predictor. METHODS: New long-sequence datasets were constructed as benchmarks for RNA Ψ-site prediction. The Ψ modification sites in RMBase 3.0 were mapped to the reference genomes across three species of human, mouse, and yeast, and the RNA sequences with a length of 201 were generated by extending the upstream and downstream from the mapped, central sites. To eliminate sequence redundancy, the sequences were clustered using CD-HIT with a 70% sequence identity threshold. We developed Meta-PseU, a logistic regression-based meta-classifier that considered 118 machine learning and deep learning classifiers. The datasets and programs are freely accessible at https://github.com/kuratahiroyuki/MetaPseU. RESULTS: By optimizing model configuration, we proposed the Meta-PseU model stacking 32 machine learning and deep learning classifiers out of 118 classifiers. Meta-PseU substantially improved model generalizability, overcoming a key limitation of existing approaches. It greatly outperformed state-of-the-art predictors and achieved increasing accuracy with increasing sequence length. CONCLUSIONS: Long-sequence datasets were newly constructed as benchmarks for RNA Ψ-site prediction. Meta-PseU offers a new framework for robust Ψ-site identification by using long sequences.

Pseudouridine↗

PubMed2026

Proteomics at scale: Bottlenecks and opportunities for early-career researchers in a fast developing field.

The field of proteomics has rapidly evolved over the last five years enabled by rapid advances in instrumentation and computation. At the same time, the proteomics community is also growing. This is reflected by the increasing participation in international conferences such as those organized by the European Proteomics Association and the Human Proteome Organization. These events provide early-career researchers with unique opportunities to exchange ideas, develop collaborations, and build networks that support professional development. One such network is the Young Proteomics Investigators Club, a European initiative supported by European Proteomics Association and led by early-career researchers. In this Community-Driven project, we investigate recent trends in proteomics by screening conference abstracts and evaluating the session attendance at Human Proteome Organization Congresses and European Proteomics Association conferences. Based on these analyses, we identified five areas that, from our perspective, are shaping the current trends in proteomics: clinical proteomics, proteomics of post-translational modifications, single-cell proteomics, systems biology and multi-omics, and computational proteomics. For each area, we highlight both unique challenges and identify a common theme: a shift from exploratory studies with manageable sample numbers towards large screenings and cohorts and the generation of big data, which often comes with the lack of computational support, organizational networks, and infrastructure. In this light, we describe the unique challenges and opportunities faced by early-career researchers. We point to actionable directions for enabling reproducible and transparent proteomics as well as community-driven projects and initiatives, which are often providing training and support. SIGNIFICANCE: In this perspective, the Young Proteomics Investigators Club (YPIC) discusses advances in analytical developments and computational approaches in proteomics research. Based on empirical analysis of recent European Proteomics Association conference and Human Proteome Organization congresses contributions, we identify clinical, single-cell, post-translational and systems-level proteomics as the research areas that have gained most momentum in the last three to five years. What makes this work distinctive is that it is written by and for early-career researchers, thereby uniquely identifying where momentum, challenges, and unmet needs converge for the newest generation of proteomics researchers. Rather than cataloguing advances, we examine the widening gap between what modern proteomics can generate and what individual researchers can realistically process, validate, and interpret. We describe specific structural barriers including access to high performance computing, limited formal training in scalable data analysis, the need for unified benchmarking standards and navigating clinical collaboration frameworks. We then highlight opportunities for the field, such as community-curated benchmarks, interdisciplinary mentorship models, and shared computational infrastructure. By making these challenges explicit from an early-career researchers standpoint, we aim to inform how training, funding, and community initiatives can be shaped to support the next generation of proteomics researchers.

Proteomics↗