Abstract While randomized controlled trials (RCTs) guide clinical practice, their completion may also influence real-world treatment patterns. We investigated whether the outcomes of trial emulation differ before and after the publication of the WARCEF (Warfarin versus Aspirin in Reduced Cardiac Ejection Fraction) randomized controlled trial. We emulated the WARCEF trial using EHR data from the Mayo Clinic Platform, comparing Warfarin and Aspirin in patients with heart failure and reduced ejection fraction (HFrEF). Analyses were stratified by the WARCEF completion date (July 2014), using intention-to-treat (ITT) frameworks. For the ITT analysis, the cohort size before 2014 consisted of 37,225 patients on Aspirin and 2022 on Warfarin, while after 2014, the cohort included 26,192 patients on Aspirin and 1080 on Warfarin. No significant treatment difference was observed before July 2014, 1.025 (95% CI: 0.6144 – 1.709, p = 0.9255), consistent with the findings of the WARCEF trial. However, after trial completion, Warfarin was associated with increased risk, 2.181 (95% CI: 1.676 – 2.839, p < 0.001). These findings initially suggested that the completion of the WARCEF trial and subsequent guideline updates were associated with changes in observed treatment effects in real-world emulation. However, further analyses indicate that the observed differences are primarily driven by the application of specific eligibility criteria rather than the trial completion date itself. In particular, the inclusion of restrictive eligibility criteria alters cohort composition and event timing, introducing selection bias that materially affects treatment effect estimates. This highlights that, in trial emulation studies, eligibility criteria can have a greater impact on observed outcomes than temporal stratification alone. Accurate emulation therefore requires careful alignment of eligibility definitions with the original trial design, as inappropriate or overly restrictive criteria may induce selection bias and distort causal inference, independent of trial publication timing.

The global rise in steatotic liver disease poses a significant public health challenge. While non-contrast computed tomography scans hold promise for opportunistic detection of steatotic liver disease, their potential for staging and risk assessment remains underexplored. Here we present a multimodal AI model trained on a large dataset, comprising of (n=968) histopathologically and (n=1103) radiologically confirmed cases, validated against both histology (n=660) and MRI-PDFF (n=375) gold standards, demonstrating high accuracy in detecting mild to severe steatosis (AUC: 0.904-0.929) and clinically significant fibrosis (AUC: 0.824-0.888). Furthermore, integrating the model into the standard clinical pathway improves primary risk screening in a retrospective patient cohort (n=1192), identifying 36% more patients at risk of fibrosis progression. Using Cox proportional hazard model, we observe that the intermediate-high risk patients identified by the optimized clinical pathway exhibits a significantly higher incidence of cirrhosis (hazard ratio: 5.54: 2.69-11.42), showcasing the model's potential for early detection and management of steatotic liver disease.

This experiment explored the impacts of potassium diformate (K-diformate) on growth performance, inflammation, intestinal barrier function, and gut microbiota in nursery piglets. Twenty-four weaned piglets were assigned to four dietary groups supplemented with 0%, 0.6%, 1.2%, or 1.8% K-diformate for 28 days. Results showed that the FCR decreased linearly with increasing K-diformate levels (p < 0.05). Importantly, piglets in the 1.8% K-diformate group had the lowest FCR, which was lower than that of the control group by 15.03% and 14.29% in weeks 1-2 and 3-4, respectively (p < 0.05). Additionally, piglets receiving 1.8% K-diformate demonstrated lower serum IL-6 levels (9.27% reduction), along with reduced jejunal IL-6 (17.64%), IL-1β (10.29%), and TNF-α levels (14.01%) relative to the control group (p < 0.05). Furthermore, piglets in the 1.8% K-diformate group exhibited decreased serum D-lactate (36.00% reduction) and LPS levels (9.90% reduction), and elevated serum GLP-2 levels (9.23% increase), as well as upregulated jejunal ZO-1, Occludin, and MUC2, both at mRNA and protein expression levels (p < 0.05). Microbiota analysis revealed piglets in the 1.8% K-diformate group exhibited higher abundances of Treponema, Roseburia, Rikenellaceae_RC9_gut_group, and norank_f_Eubacterium_coprostanoligenes_group, while showing lower abundances of Lactobacillus, Streptococcus, Sarcina, and Blautia (p < 0.05). Correlation analysis further revealed associations between Roseburia, Blautia, Sarcina, and Lactobacillus with inflammation, and between Treponema, Roseburia, Streptococcus, Blautia, and Sarcina with intestinal barrier function (p < 0.05). Collectively, dietary addition with K-diformate (especially 1.8%) improved growth performance, inflammation, and intestinal barrier function associated with gut microbiota modulation in nursery piglets.

Coarsening Visium HD resolution from 8 to 64 μm can flip cell-type colocalization from negative to positive ( r = - 0.12 → + 0.80 ) , yet many widely used compositional deconvolution workflows require coarsening or subsampling at million-bin scale. Here we introduce FlashDeconv, which combines leverage-score importance sampling with sparse spatial regularization to achieve competitive benchmark accuracy while processing 1.6 million bins in 153 seconds on commodity hardware. Systematic multi-resolution analysis of Visium HD mouse intestine reveals a tissue-specific resolution horizon (8-16 μ m)-the scale at which this sign inversion occurs-validated by Xenium ground truth. Below this horizon, FlashDeconv provides, to our knowledge, the first sequencing-based quantification of Tuft cell chemosensory niches (15.3-fold stem cell enrichment). In a 1.6-million-bin human colorectal cancer cohort, FlashDeconv uncovers neutrophil inflammatory microdomains co-localized with immunoregulatory dendritic cells (mRegDC) at the tumor-stroma interface-spatial niches largely missed by discrete-label summaries, with RCTD doublet mode labeling only 2.3% of hotspot bins as neutrophil singlets.

Breast cancer (BC) remains a leading cause of cancer-related mortality worldwide, and accumulating evidence suggests that tumor-associated microbiota may contribute to disease heterogeneity beyond host genetic and immune determinants. Advances in sequencing and multi-omics technologies have uncovered a reproducible intratumoral microbiome in BC, with distinct compositional patterns associated with molecular subtypes, clinicopathological features, and clinical outcomes. Alterations in specific microbial taxa have also been linked to tumor immune status, metastatic potential, and therapeutic sensitivity, underscoring their potential value in disease stratification and prognostic assessment. Although breast tissue represents a low-biomass environment, multiple studies employing stringent contamination control strategies have confirmed the reliability of these microbial signals. Experimental evidence further demonstrates that intratumoral microbes are functionally active components of the tumor microenvironment (TME), influencing tumor progression and metastasis through immune modulation, inflammatory signaling, and metabolic or hormonal reprogramming, while also shaping responses to chemotherapy and immunotherapy. This review summarizes current knowledge on the compositional features, functional mechanisms, and clinical relevance of the BC intratumoral microbiome, highlights methodological challenges in low-biomass profiling, and discusses future directions for translating these findings into clinically actionable strategies. The aim of this review is to systematically evaluate the role of the intratumoral microbiome in breast cancer pathogenesis and treatment, and to propose a framework for translating current findings into clinical practice.

The rapid expansion of single-cell RNA sequencing (scRNA-seq) has made accurate cell type annotation a critical bottleneck for biological discovery. Existing computational methods are often limited by reference data dependency, while emerging single Large Language Model (LLM) approaches are susceptible to model-specific biases and provide insufficient uncertainty quantification. To address these limitations, we introduce mLLMCelltype, a framework that harnesses collective intelligence-the emergent problem-solving capacity arising when multiple independent agents interact through structured deliberation to produce solutions exceeding individual capabilities-of multiple LLMs through an iterative deliberation process. Across 49 diverse datasets, our framework achieves a mean accuracy of 77.2%, a 15.7-percentage-point improvement over the best-performing single-LLM baseline (61.5%). The consensus mechanism demonstrates high robustness to noisy input and generalizes to datasets released after the LLMs' training. By providing transparent reasoning chains and robust consensus-based confidence metrics, mLLMCelltype minimizes manual annotation effort and enables reliable interpretation of complex cellular landscapes. The framework is available as an open-source package and an accessible web server.

Spatial transcriptomics has transformed our ability to study tissue architecture at molecular resolution, yet analyzing these data demands navigating dozens of computational methods across incompatible Python and R ecosystems-forcing researchers to devote more effort to making tools function than to pursuing biological questions. We present ChatSpatial, a platform in which the LLM selects from pre-validated tool schemas rather than generating free-form code, with domain expertise embedded in schema descriptions for context-aware parameter inference. Built on the Model Context Protocol (MCP), ChatSpatial unifies 60+ methods across 15 analytical categories into a single conversational workflow spanning Python and R ecosystems. Replication of two published studies-recovering subclonal heterogeneity in ovarian cancer and tumor microenvironment organization in oral squamous cell carcinoma-and validation across seven LLM platforms demonstrate that schema-enforced orchestration yields near-deterministic reproducibility at the workflow level for multi-step spatial analyses. Beyond replication, exploratory cross-method analyses illustrate practical triangulation across independent analytical frameworks.

AIMS:Metabolic dysfunction-associated steatotic liver disease (MASLD) is increasingly linked to cognitive decline, yet the hepatic factors associated with imaging-derived neurovascular coupling (NVC) remain unclear. This study aimed to investigate whether liver stiffness or liver fat content was more closely associated with resting-state CBF-ReHo surrogate measures. MATERIALS AND METHODS:A total of 130 participants including 98 MASLD patients and 32 age- and education-matched healthy controls (HCs) underwent clinical assessment, neuropsychological testing, and multi-modal MRI. Liver stiffness and fat content were quantified using MR elastography (MRE) and MRI-proton density fat fraction (PDFF). Imaging-derived NVC was assessed using global cerebral blood flow (CBF)-regional homogeneity (ReHo) coupling and voxel-wise CBF/ReHo ratios, interpreted as resting-state surrogates rather than direct stimulus-evoked NVC. Multivariable regression and exploratory mediation analyses were employed. RESULTS:Compared to HCs and patients with lower liver stiffness (MASLD_low), those with higher liver stiffness (MASLD_high) showed reduced global CBF-ReHo coupling and altered CBF/ReHo ratios, primarily localized to the bilateral superior temporal pole/superior temporal gyrus (TPOsup). In multivariable regression, liver stiffness remained independently associated with TPOsup CBF/ReHo ratios (p < 0.001). Exploratory mediation analysis showed a statistically significant indirect association between hepatocellular injury markers and TPOsup CBF/ReHo ratios involving MRE-derived liver stiffness (95% CI: 0.0795-0.2790). CONCLUSIONS:Within this cohort and the observed PDFF range, MRE-derived liver stiffness was more closely associated with imaging-derived NVC surrogate measures than liver fat content in MASLD. These findings are hypothesis-generating and require validation in longitudinal studies.

Methods that map genetic risk to cells do not directly test whether spatially organized ligand-receptor (LR) gene annotations carry conditional heritability association. Here we introduce EdgeMap, which scores each gene by the spatial activity of its LR contexts in spatial transcriptomics data, maps these scores alongside cell-intrinsic annotations to SNP-level LD scores, and tests both jointly against GWAS summary statistics. Across 17 traits and five human tissues, edge z -scores are systematically higher in biologically matched tissues: all 15 traits with both matched and unmatched tissue classes show this ordering (median Δ z = 1.51 ; sign P = 3.1 × 10 -5 ), and 11 of 13 nominal-positive associations concentrate in the biologically matched set ( P = 1 × 10 -4 ). The pattern survives broad LR-gene-class and cell-type composition controls. Donor-level cross-section summaries, independent GWAS replication for LDL and CAD, and cell-segmented Visium HD liver data provide convergent support. A secondary analysis prioritizes constituent genes within active LR contexts; sixteen of 31 prioritized genes are absent from standard gene-level methods. These results establish spatially weighted LR-gene annotations as a complementary layer for interpreting complex-trait genetic architecture.

Identifying spatially variable genes (SVGs) is the first analytical step in spatial transcriptomics, determining which genes and pathways are prioritized for downstream validation. Yet the restricted spatial models of current detection methods create systematic blind spots that can exclude biologically coherent programs from discovery. Here we present FlashS, which reformulates kernel-based spatial testing in the frequency domain to detect arbitrary multi-scale expression patterns while scaling to millions of cells. In human cardiac tissue, this broader detection capacity recovers a coherent PGC-1α-regulated mitochondrial biogenesis program-40 of 49 pathway genes spatially associated with ventricular cardiomyocytes-that PreTSA, a leading parametric alternative, largely misses (1 of 49 genes), a finding replicated in an independent cohort. Across 50 benchmark datasets spanning 9 platforms, FlashS achieves state-of-the-art ranking accuracy (mean Kendall τ = 0.935 ) and completes on the Allen Brain MERFISH atlas (3.94 million cells) in 12.6 minutes with 21.5 GB memory.

Power analysis is a critical step in designing a microbiome study. Existing power calculation tools for microbiome studies mainly rely on parametric models of the sequencing counts, which underestimate the complexity of microbiome data and could produce overly optimistic power estimates. In this work, we present a new simulation-based power analysis tool, mPower, for microbiome study design. The tool uses a real data-based semi-parametric simulation framework to generate realistic microbiome data, upon which the power assessment is performed. Coupled with a select differential analysis tool, our power tool supports different study designs, including cross-sectional, case-control, and matched-pair studies, with or without confounders. It allows power analysis for both community-level and taxon-level testing. By using microbiome reference datasets from different environments, the users could perform power calculation based on the environment of interest. The mPower is primarily designed for 16S amplicon sequencing data, and it also incorporates a parametric simulation framework that enables power analysis for shotgun metagenomic data. We showcase the application of mPower with several real-world examples. The web interface of mPower is available at https://microbiomestat.shinyapps.io/mPower/.

BACKGROUND & AIMS:Obesity is a heterogeneous disease characterized by different pathophysiological and behavioral traits that influence response to glucagon-like peptide 1 (GLP-1)-based therapies. We previously identified an obesity phenotype characterized by fast gastric emptying (GE) and increased postprandial hunger. We aimed to elucidate pathophysiological mechanisms in this phenotype by evaluating plasma enteroendocrine hormones and mucosal gene expression and to evaluate treatment response to tirzepatide across subphenotypes. METHODS:A total of 483 adults with obesity underwent solid meal GE (SGE) by scintigraphy, postprandial appetite assessment using a visual analogue scale, and plasma enteroendocrine hormone profiling. Gaussian mixed modeling identified phenotypic clusters. Associations with plasma short-chain fatty acids and fecal metagenomics were explored. A separate cohort (n = 31) underwent colonic mucosal biopsies with quantification of GCG (GLP-1) and PYY messenger RNA. Retrospective evaluation of weight loss in participants treated with tirzepatide among each cluster was performed (n = 61). RESULTS:Three clusters were identified based on SGE and GLP-1. One cluster demonstrated fast SGE, increased postprandial hunger, and discordantly low postprandial GLP-1 (termed dc-GE/GLP-1; n = 130 [26.9%]), as well as lower plasma peptide YY and cholecystokinin. dc-GE/GLP-1 showed higher plasma short-chain fatty acid levels, without significant differences in fecal microbial composition. Compared with concordant clusters (c-GE/GLP-1; n = 353 [73.1%]), dc-GE/GLP-1 had decreased mucosal messenger RNA expression of GCG (GLP-1) and PYY. At 6 months of tirzepatide, dc-GE/GLP-1 was associated with greater weight loss compared with c-GE/GLP-1 (21.5% vs 11.7%). CONCLUSIONS:We identified a subphenotype of obesity with fast GE and discordantly low GLP-1 plasma levels, reduced mucosal hormone synthesis, and enhanced weight loss to tirzepatide. Further studies are needed to identify mechanisms contributing to GLP-1 deficiency in this subphenotype of obesity.

Purpose: To establish normal myocardial stiffness values in healthy adults using cardiac MR elastography (MRE) and assess the impact of slice location and demographic factors. Methods: 125 healthy volunteers underwent cardiac MRE using a 140-Hz vibration frequency, with data acquired during early systole (approximately 100 ms after the R-wave). Stiffness was measured in basal, mid, and apical short-axis views. Statistics included repeated-measures ANOVA (slice locations), t-tests (sex), Spearman correlation (age), and multivariable linear regression (controlling for age, sex, body mass index, and body surface area). Results: Global myocardial stiffness was 8.1 f 1.1 kPa. A significant apical-to-basal gradient existed (apex: 8.7 f 1.1, mid: 8.3 f 1.2, base: 7.4 f 1.3 kPa; all P < 0.05). Males had non-significantly higher global stiffness than females (8.3 f 1.0 vs. 8.0 f 1.1 kPa, P = 0.068). Stiffness correlated with age in both sexes (r = 0.48 in male, r = 0.42 in female, both P < 0.05). Multivariable regression confirmed age as an independent predictor (beta = 0.037 kPa/year, P < 0.001) with no age-sex interaction (P = 0.929).Intra-/inter-observer and scan-rescan reproducibility were excellent (ICCs >= 0.97). Conclusion: This study establishes that myocardial stiffness in healthy adults follows an apical-to-basal gradient and is independently influenced by age, providing location-specific baseline stiffness values for the clinical interpretation of cardiac MR elastography. delete this paragraph-Advances in knowledge: This study establishes comprehensive, location-specific normal ranges for myocardial stiffness across the entire left ventricle (apex, mid, base) in healthy adults, addressing a key gap in the existing CMRE literature which has been largely confined to mid-ventricular data. It demonstrates the significant impact of slice location on stiffness values, revealing an apical-to-basal gradient. These findings underscore the necessity of specifying and standardizing slice location in future CMRE protocols for accurate interpretation and comparison, both in research and potential clinical applications. The study also provides a robust, confounder-adjusted analysis confirming the independent effect of age on myocardial stiffness.

9573 Background: Clinical Stage III (cSIII) melanoma patients are at high risk of recurrence following therapeutic lymph node dissection (TLND) and adjuvant therapy. Neoadjuvant immunotherapy based regimens have shown promising results in this setting, with pathological response being the best predictor of recurrence free survival (RFS) and distant metastases free survival (DMFS). T-cell Receptor (TCR) abundance and clonality are being explored as biomarkers. Methods: In the phase II NeoACTIVATE (NCT03554083) trial, patients were treated based on BRAF status with 12 weeks of neoadjuvant atezolizumab, vemurafenib and cobimetinib (Arm A, BRAF-mutated) or atezolizumab and cobimetinib (Arm B, BRAF wildtype), followed by TLND and 6 months adjuvant atezolizumab. TCR sequencing (n=23) was performed on the pretreatment involved lymph node (iLN) and on the same node, along with an adjacent uninvolved node (uiLN) after TLND. The fraction productive of T cells (fpTC) was defined as the percentage of in-frame TCR sequences (capable of encoding a functional expressed receptor protein) of all TCR sequences identified. Simpson Clonality Index (SCI) was used to measure relative abundance of specific clones, with a higher score indicating a more monoclonal repertoire. Comparisons were made using the Wilcoxon Rank Sum test. Results: 30 patients were enrolled, 15 in each arm. 2 patients were not operated on, and one did not receive adjuvant treatment, leaving 27 patients for outcome analysis. 3-year RFS was 55.6% (95% CI 9.5%-77.8%), median of 62.2 months (95% CI 19-NR); DMFS was 64.3% (95% CI 48%-86.3%), median 62.2 months (95% CI 33.9-NR). The fpTC was higher in the pretreatment iLN for those without an RFS event (p=0.04) with a trend in the uiLN as well (p=0.06). Patients without a DMFS event had a higher fpTC in the uiLN (p=0.03) but not in the iLN. The SCI was significantly higher only in the post treatment iLN for patients without an RFS event (p=0.03) and was not significantly different when compared based on DMFS status. When performing pairwise comparisons of nodes based on recurrence status, we found that patients who did not suffer an RFS event had a significantly higher SCI in the post treatment iLN compared to the uiLN (p=0.02). Conclusions: Neoadjuvant immunotherapy and targeted therapy for cSIII melanoma can lead to durable RFS. Higher pretreatment fpTC as well as higher post-treatment iLN TCR clonality were both associated with improved RFS. These data indicate that TCR repertoire characteristics within iLN and uiLN may reflect differential immune responses to neoadjuvant therapy and could inform future studies evaluating biomarkers to stratify risk of recurrence and allow personalization of therapy. Clinical trial information: NCT03554083 . fpTC Simpson Clonality Index RFS event DMFS event RFS event DMFS event Pretreatment iLN 0.04 0.12 0.43 0.11 Posttreatment iLN 0.09 0.09 0.03 0.12 uiLN 0.06 0.03 0.79 0.48 Δ iLN (post-pre) 0.79 0.37 0.07 0.41 Δ posttreatment iLN – uiLN 0.21 0.21 0.02 0.08

BACKGROUND:Long COVID (LC) manifests in 10%-30% of non-hospitalized individuals post-SARS-CoV-2 infection, leading to significant morbidity. The predictive role of gut microbiome composition during acute infection in the development of LC is not well understood, partly because of the heterogeneous nature of the disease. OBJECTIVES:To determine whether the gut microbiome composition in the acute phase of SARS-CoV-2 infection predicts subsequent LC and to investigate the role of microbiome signatures in disease subphenotypes. DESIGN:We conducted a longitudinal cohort study involving 799 outpatient participants tested for SARS-CoV-2 due to similar symptom presentation, including 380 SARS-CoV-2 positive and 419 negative individuals. Stool samples were collected at two time points for metagenomic sequencing. Logistic regression with L1 regularization was employed to predict LC based on the microbiome and clinical metadata. RESULTS:The individuals who developed LC harbored a distinct gut microbiome during acute infection compared to those who recovered fully and uninfected controls with similar symptomatology. However, the temporal changes in the gut microbiome between the acute (0-1 month) and post-acute (1-2 months) phases were similar across the three cohorts. Using machine learning, we showed that the gut microbiome carried a modest signal for subsequent LC, but model performance was insufficient for clinical prediction, likely reflecting the heterogeneous nature of LC. Finally, we identified four LC symptom clusters, with gastrointestinal and fatigue-only groups strongly linked to gut microbiome alterations. CONCLUSION:The gut microbiome can potentially offer solutions for understanding the heterogeneous nature of LC. Larger cohorts and phenotype-aware computational algorithms may help overcome current model performance limitations and support the development of targeted diagnostic and therapeutic strategies.

RATIONALE AND OBJECTIVES:To characterize liver fat distribution in metabolic dysfunction-associated steatohepatitis (MASH) and propose a magnetic resonance imaging proton density fat fraction (MRI-PDFF)-based score for MASH identification in obesity. MATERIALS AND METHODS:Individuals with obesity were recruited for liver biopsy and contemporaneous MR scanning. The variability in fat distribution was evaluated by calculating the standard deviation (SD), range, and coefficient of variation of MRI-PDFF across liver lobes. A machine learning-assisted strategy was employed to establish a diagnostic model for MASH. RESULTS:A total of 107 participants with biopsy-confirmed fatty liver were included, and 44 were diagnosed with MASH. The MASH group exhibited significantly higher fat content in all liver segments and more homogeneous fat distribution in the right liver lobe than those with non-MASH. A raw-MASH score, incorporating the mean value and SD of PDFF in the right lobe, alanine aminotransferase levels, and waist circumference, was then established to identify MASH with an area under the receiver operating characteristic curve (AUROC) of 0.90 (95%CI 0.84-0.95). It performed better than HAIR, ION, and acNASH (all p < 0.05), and exhibited higher AUROC than MR-MASH model (0.90 vs. 0.85, p = 0.066). Besides, the dual threshold strategy of raw-MASH improved the diagnostic performance with high sensitivity and specificity. CONCLUSION:In addition to the hepatic fat content, the homogeneity of fat distribution may represent another significant hallmark of MASH. The raw-MASH score, which is available from MR imaging and routine clinical collection, shows great potential for identifying MASH in obesity.

Abstract Background: Aberrations in telomere length have important implications in cancer development and progression. This study aimed to determine whether leukocyte telomere length (LTL) in patients with colorectal cancer (CRC) is associated with survival outcomes. We also investigated whether genetic variants in telomere maintenance genes are associated with survival in these patients. Methods: Blood specimens were collected from 1,007 patients prior to receiving chemotherapy or radiation. Genomic DNA was extracted using the Promega Maxwell RSC instrument. LTL was measured in triplicate using monochrome multiplex PCR, where the average amplification value of telomeric repeats was termed T. Similarly, PCR amplification of a single-copy reference gene was performed, and its average amplification value was termed S. The telomere length for each sample was expressed as the T/S ratio. Genotyping of single-nucleotide polymorphisms (SNPs) in TERC, TERT, and OBFC1 genes was conducted at the institutional Genome Analysis Core facility. Kaplan-Meier survival curves were generated to evaluate patient survival outcomes. Results: Younger individuals (∼25 years) exhibited nearly a twofold longer LTL compared with older individuals (∼75 years). No significant difference in LTL was observed between stage II and stage III CRC patients. A strong inverse correlation was observed between patient age and LTL (Spearman’s r = -0.48; p = 1.13 × 10-58). Females had significantly longer LTL than males (p = 3.97 × 10-5). The TERC SNP rs1317082 was significantly associated with both overall survival (OS) and disease-free survival (DFS) in the combined stage II and stage III patient cohort (p = 0.017 and p = 0.023, respectively). Similarly, the OBFC1 SNP rs9419958 was significantly associated with OS (p = 0.016). Importantly, LTL itself was significantly associated with both OS and DFS (p = 0.008 and p = 0.044, respectively) among the combined stage II and stage III patients. Kaplan-Meier survival analyses demonstrated that age- and sex-adjusted LTL was predictive of long-term survival outcomes. Conclusions: Survival among patients with stage II and III colorectal cancer is significantly influenced by LTL. Additionally, specific allelic variants of telomere maintenance genes, including TERC and OBFC1, are independently associated with improved survival, irrespective of LTL. These findings suggest that peripheral blood LTL measurement, along with telomere-related genotyping, may serve as a valuable prognostic marker for colorectal cancer outcomes. Funding: Individualizing colorectal cancer patient care using the host and tumor telomere phenotype (RO1 CA204013), Curtiss Fund (92541775), C-SiG Core(s): Epigenomics & Spatial Biology Core, and Clinical Core of the Mayo Clinic Center for Cell Signaling in Gastroenterology (P30DK084567) Citation Format: Estela M. Cruz Garcia, Gobinda Sarkar, Jun Chen, Shubham Sood, Kim Kossick, Daniel Schupack, Rondell Graham, Brooke Druliner, Zahra Heydari, Lauren Helgeson, Richard G. Cawthon, Lisa A. Boardman. Longer leukocyte telomeres and specific alleles of telomere maintenance genes are independently associated with improved survival of colon cancer patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1919.

Objective:To develop and validate a CT-based habitat imaging model incorporating intratumoral microenvironment heterogeneity analysis for noninvasive preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC), addressing limitations of conventional radiomics in characterizing intra-tumor heterogeneity. Methods:This retrospective study included 216 patients with pathologically confirmed HCC undergoing resection. Preoperative portal venous phase contrast-enhanced CT and corresponding postoperative histopathologic MVI status were collected. Habitat (functional heterogeneous subregion) and conventional radiomic features were extracted. A habitat risk score and radiomic score were calculated. Clinical-pathologic factors (Edmondson grade, p53/CD10 expression, tumor diameter) identified via univariate/multivariate analyzes were integrated into a predictive model visualized as a nomogram. Performance was assessed by AUC, calibration curves, and decision curve analysis (DCA). DeLong's test compared ROC curves. Results:In the training set, the combined model achieved an AUC of 0.862 (95% CI: 0.797-0.926); in the validation set, AUC was 0.814 (95% CI: 0.710-0.918), both significantly outperforming individual models (all P < 0.05). Calibration curves showed good agreement (Hosmer-Lemeshow P = 0.60). DCA indicated net benefit at thresholds of 15%-65%. DeLong's test confirmed the combined model had higher AUC than the clinical model (Z = -3.21, P < .05) and radiomics-only model (Z = -2.05, P < .05). Conclusion:The CT-based habitat imaging model quantified intratumoral heterogeneity and, combined with clinicopathologic data, provided a reliable noninvasive tool for preoperative MVI risk stratification in HCC, with strong clinical potential.