Citations

Abstract With the rapid advancements in large language model technology and the emergence of bioinformatics‐specific language models (BioLMs), there is a growing need for a comprehensive analysis of the current landscape, computational characteristics, and diverse applications. This survey aims to address this need by providing a thorough review of BioLMs, focusing on their evolution, classification, and distinguishing features, alongside a detailed examination of training methodologies, datasets, and evaluation frameworks. We explore the wide‐ranging applications of BioLMs in critical areas such as disease diagnosis, drug discovery, and vaccine development, highlighting their impact and transformative potential in bioinformatics. We identify key challenges and limitations inherent in BioLMs, including data privacy and security concerns, interpretability issues, biases in training data and model outputs, and domain adaptation complexities. Finally, we highlight emerging trends and future directions, offering valuable insights to guide researchers and clinicians toward advancing BioLMs for increasingly sophisticated biological and clinical applications.

Abstract Objective: The Pancreatic Cancer Detection Consortium (PCDC) performed a blinded EDRN defined phase 2 Biomarker Bakeoff study of eight promising biomarker panels using pancreatic cancer ductal adenocarcinoma (PDAC) case and general population healthy control samples from three institutions. The scientific questions were to determine whether panels could discriminate PDAC cases from controls with Receiver Operating Characteristic Area Under the Curve (AUC) > 0.8, compare panel performance head-to-head via 95% confidence intervals (CI), and determine whether the panel improved upon serum CA19-9 alone. A secondary goal was to perform modeling to discover a new panel consisting of individual biomarkers. Methods: Eligible biomarkers were required to have had a prior EDRN defined phase 2 validation study published and be ready for validation in a sample set from multiple institutions. Three institutions submitted a total of 140 cases and 140 controls frequency matched to cases on age and sex. Four consortium member institutions submitted for validation eight panels consisting of ten plasma, serum and DNA genotyping biomarkers in addition to CA19-9. Locked panel formulas were submitted prior to specimen distribution. The primary metric was AUC with bootstrapped 95% CIs. LASSO penalized regression was used for exploratory panel building. Nested cross validation and internal-external validation were performed, and optimism-corrected performance estimates were computed. Results: The overall sample (N=280) was 51% female, median age of 67 years, and 94% were of non-Hispanic white race and ethnicity. Controls had median BMI 27.3, 56% never smokers, and 12% type II diabetes mellitus (DM2); cases (staged I to IV) had median BMI 26.6, 42% never smokers, and 26% DM2. Overall panel AUC estimates ranged from 89.9 to 96.3 and all were significantly larger than the null hypothesis value of 0.8 (p<0.05). The apparent serum CA19-9 AUC was 91.7 (87.8, 95.6). Two panels were significantly different from serum CA19-9, a panel with FUT2/3 genotype with serum CA19-9 (p=0.002) and a panel containing tissue factor pathway inhibitor (TFPI), tenascin C (TNC), and plasma CA19-9 (p=0.01). In subset analyses by PDAC stage, no panel improved upon the null hypothesis value of 0.8 in stage I (p>0.05), all panels improved upon the null hypothesis value of 0.8 in stage II and III (p<0.05), and only serum CA19-9 did not improve upon the null hypothesis value of 0.8 in stage IV. All biomarkers but two microRNAs were retained in a LASSO model that had optimism corrected AUC=96.4 (95% CI: 94.0, 98.9). Conclusions: All submitted panels performed well overall in this blinded study, with the panel containing FUT2/3 genotype performing the best. Panel discovery improved upon serum CA19-9, though optimism corrected AUC 95% CIs overlap with those of apparent CA19-9 performance. Further work is needed to prioritize promising biomarkers and to improve detection of early stage PDAC. Citation Format: Ann L. Oberg, William R. Bamlet, Grant Izmirlian, Seetharaman Balasenthil, Surinder K. Batra, Masataka Hayashi, Daniel Herman, Michael A. Hollingsworth, Maneesh Jain, Ann M. Killary, Brianna M. Krusen, Suyu Liu, Gopalakrishnan Natarajan, Subrata Sen, Lynette M. Smith, Sudhir Srivastava, Brian M. Wolpin, Kenneth S. Zaret, Michael G. Goggins. Pancreatic Cancer Detection Consortium biomarker bake-off: A consortium Phase 2 blinded biomarker validation and panel discovery study [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pancreatic Cancer Research—Emerging Science Driving Transformative Solutions; Boston, MA; 2025 Sep 28-Oct 1; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(18_Suppl_3):Abstract nr B070.

Abstract Pancreatic adenocarcinoma (PDA) has poor response to chemotherapy with a strong likelihood of cancer recurrence. While previous studies have examined patient samples collected before and after chemotherapy, few have analyzed matched samples from the same patients that allow us to deconvolute interpatient heterogeneity, a large limitation to PDA studies. Here, we study the effects of chemotherapy on tumor cells and their surrounding microenvironment using scRNA sequencing of matched pre- and post- treatment tissue biopsies from 5 patients. We utilized the Seurat v4 pipeline to observe transcriptomic differences and corresponding putative cellular interactions in both treatment states. Unsurprisingly, we found insufficient tumor epithelial cells post-treatment due to the intended cytotoxic effects of chemotherapy, leading us to focus on the tumor microenvironment where we had sufficient cells to analyze. We identified that immune cells in the TME are heterogeneously affected by chemotherapy treatment with patient specific differences, but we did identify a unanimous increase in CXCR4 in CD4 and CD8 positive T cells after treatment. Analysis of CAFs revealed that a subset of fibroblasts lose myofibroblastic features and increasingly acquire 'neural'-like features after treatment. This suggests that chemotherapy-altered fibroblasts may contribute to a pro-tumor microenvironment that facilitates recurrence through neuron–CAF–tumor interactions. Through inferred ligand-receptor pairs identified with the CellChat v2 pipeline, we observed that increased CXCL12 in these neural CAFs interacts with the increased CXCR4 on T-cells. We validated the expression of these genes in patients by staining human patient PDAC tissues (both treated and treatment naïve tissues) for neural genes that were highly enriched post-treatment in our single cell sequencing dataset. We have also begun inhibiting these genes of interest in CAFs and are exploring the effects of these ‘neural’-CAFs on patient-derived tumor organoids in culture. Citation Format: Aylin Z. Henstridge, Padma Kadiyala, Ahmed Elhoissiny, Jianhua Liu, Georgina Branch, Nandini Arya, Vaibhav Sahai, Nicole Peterson, Jorge Machicado, Richard Kwon, Allison Schulman, Erik Wamsteker, George Phillips, Martin Fernandez-Zapico, Mark Truty, Timothy Frankel, Fillip Bednar, Marina Pasca di Magliano, Costas Lyssiotis, Eileen S. Carpenter. An increase in neural cancer associated fibroblasts following exposure to chemotherapy in pancreatic ductal adenocarcinoma [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pancreatic Cancer Research—Emerging Science Driving Transformative Solutions; Boston, MA; 2025 Sep 28-Oct 1; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(18_Suppl_3):Abstract nr B079.

Background The eccentric stenosis morphology in patients with intracranial atherosclerotic stenosis (ICAS) has a critical impact on the efficacy of endovascular treatment. This study aimed to establish a rabbit model of eccentric carotid atherosclerotic stenosis that mimics the pathophysiological and morphological characteristics of eccentric plaques in humans. Methods Rabbits received localized cryoinjury of the right carotid artery after 2 weeks of a high fat, high cholesterol diet, followed by an additional 8 weeks on the same diet. On the experimental endpoint, comprehensive analyses of the model’s pathological changes were conducted using serum lipid analysis, DSA, optical coherence tomography (OCT), histopathological analysis, and immunohistochemistry. Results The high fat, high cholesterol diet significantly elevated serum lipid levels in rabbits. Hematoxylin and eosin staining results at 24 hours postoperatively revealed focal intimal damage induced by 5 s of cryoinjury. After localized cryoinjury and high fat, high cholesterol diet feeding, DSA results demonstrated significant carotid stenosis, with a stenosis degree of 47±5% and a length of approximately 9 mm. OCT images revealed eccentric fibrous plaque formation at the stenotic site. Morphological analysis of the plaque samples demonstrated a plaque eccentricity index of 0.66±0.05 and a plaque burden of 60.68±7.41%. Histopathological analysis showed that atherosclerotic lesions developed in all rabbits of the model group. Among them, three rabbits had intermediate lesions, one rabbit had atheroma, and eight rabbits had fibroatheroma. Conclusions This study successfully established a rabbit model of eccentric carotid atherosclerotic stenosis, providing a novel experimental platform for transformation research of neurointerventional devices.

Abstract Background Delirium is often underdiagnosed in clinical practice and is not routinely coded for billing. While manual chart review can identify delirium, it is labor-intensive and impractical for large-scale studies. Natural language processing (NLP) can analyze unstructured text in electronic health records (EHRs) to extract meaningful clinical information. Methods To support national integration of NLP for EHR-based delirium identification across different institutions, we launched the Delirium Interest Group within the national Evolve to Next-Gen Accrual to Clinical Trials (ENACT) NLP Working Group. This paper outlines our initial efforts to standardize, evaluate, and translate an NLP-based delirium detection model into the i2b2/ENACT platform. Results Multisite contextual inquiry identified several key challenges, including variations in local screening practices (eg, tools used, documentation frequency, and quality control), the need for harmonized definitions in the context of EHRs, and the complexity of modeling temporal logic. Multisite NLP evaluation revealed variable performance degradation driven by differences in delirium screening practices, clinical documentation patterns and semantics, and note syntactic structures. Conclusion Our work represents an important first step toward enabling scalable and standardized NLP-based delirium detection across institutions. By engaging diverse institutions through the ENACT NLP Working Group, we identified shared challenges and site-specific variations that impact model implementation and performance. Our collaborative approach enabled the development of a more robust framework for delirium identification across heterogeneous EHR systems. Future efforts will build on this foundation to enhance the validity, usability, and translational impact of delirium detection.

PURPOSE Large language models (LLMs) have demonstrated remarkable versatility in oncology applications, such as cancer staging and survival analysis. Despite their potential, ethical concerns such as data privacy breaches, bias in training data, lack of transparency, and risks associated with erroneous outputs pose significant challenges to their adoption in high-stakes oncology settings. Therefore, we aim to explore the ethical challenges associated with LLM-based applications in oncology and evaluate emerging techniques designed to address these issues. METHODS Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework, a systematic review was conducted to evaluate publications related to ethical issues of LLMs in oncology across eight academic databases (eg, PubMed, Web of Science, and Embase) between January 1, 2019, and December 31, 2024. RESULTS The search retrieved 4,319 published articles, of which 65 publications were preserved and included in our analysis. We identified six prevalent ethical challenges in oncology, including trust, equity, privacy, transparency, nonmaleficence, and accountability. We then evaluated emerging technical solutions to mitigate ethical challenges and summarized evaluation metrics used to assess these solutions' effectiveness. CONCLUSION This review provides actionable recommendations for responsibly deploying LLMs in oncology, ensuring adherence to ethical guidelines, and fostering improved patient outcomes. By bridging technical and clinical perspectives, this review offers a foundational framework for advancing ethical artificial intelligence applications in oncology and highlights areas for future research.

Objective To establish a framework for validating candidate biomarkers of cerebral small vessel diseases (SVD) associated with cognitive impairment and characterize individuals enrolled by the MarkVCID2 consortium under this framework. Methods Participants age 60 to 90 years were enrolled across 17 MarkVCID2 sites. Recruitment was targeted to enrich in cognitive symptoms (mild dementia, mild cognitive impairment, subjective cognitive decline), defined risk factors (diabetes mellitus, advanced hypertension), and Black/African American, White, and Hispanic/Latino subgroups. Enrolled participants underwent baseline visits that included cognitive testing, multimodal magnetic resonance imaging (MRI), and biofluid collection. Provisional risk for SVD‐related cognitive decline was estimated primarily by baseline cognitive symptoms plus SVD risk factors. Adjudicated risk status was estimated by cognitive symptoms plus presence of moderate‐to‐severe white matter hyperintensities, microbleeds, or lacunes on baseline MRI. Results MarkVCID2 enrolled 1883 individuals age 73.4 ± 7.5 years, 65.0% female, 24.2% Hispanic, and 27.1% non‐Hispanic Black. Among enrollees, 44.8% were provisionally designated high‐risk. After baseline MRI, 48.5% were categorized as adjudicated high‐risk status, with substantial recategorization both from low‐ to high‐risk (primarily because of MRI lesions without SVD risk factors) and high‐ to low‐risk (primarily suspected cognitive impairment at screening not confirmed by baseline testing). Interpretation MarkVCID2 baseline data indicate successful enrollment of diverse individuals enriched in factors associated with SVD‐related cognitive decline. Changes over 3 years of longitudinal follow‐up will be analyzed to validate the candidate biomarkers for 2 projected contexts of use: subject selection (identifying likelihood of future SVD progression) and study outcome (efficiently measuring SVD progression). ANN NEUROL 2026;99:449–458

Abstract Background Lung adenocarcinoma in never-smokers is a major public health burden, especially among East Asian women. Polygenic risk scores (PRSs) are promising for risk stratification but are primarily developed in European-ancestry populations. We aimed to develop and validate single- and multi-ancestry PRSs for East Asian never-smokers to improve lung adenocarcinoma risk prediction. Methods PRSs were developed using genome-wide association study summary statistics from East Asian (8002 cases; 20 782 controls) and European (2058 cases; 5575 controls) populations. Single-ancestry models included a PRS with 25 single nucleotide variants (PRS-25), PRS clumping and thresholding (CT), and LDpred2; multi-ancestry models included LDpred2 plus European PRS-128, PRS-CSx, and clumping and thresholding, super learning, empirical Bayes (CT-SLEB). Performance was evaluated in independent East Asian data from the Female Lung Cancer Consortium and externally validated in the Nanjing Lung Cancer Cohort. We assessed predictive accuracy via area under the curve operating procedure (AUC), with 10-year and (ages 30-80 years) absolute risks estimates. Results The best multi-ancestry PRS, using East Asian and European data via CT-SLEB, outperformed the best East Asian-only (LDpred2; AUC = 0.629, 95% confidence interval [CI] = 0.618 to 0.641), achieving an AUC of 0.640 (95% CI = 0.629 to 0.653) and odds ratio of 1.71 (95% CI = 1.61 to 1.82) per standard deviation increase. Nanjing Lung Cancer Cohort validation confirmed robust performance (AUC = 0.649, 95% CI = 0.623 to 0.676). The top 20% PRS group had a 3.92-fold higher lung adenocarcinoma risk than the bottom 20%. Further, the top 5% PRS group reached a 6.69% lifetime absolute risk. Notably, this group reached the average population 10-year lung adenocarcinoma risk at age 50 years (0.42%) by age 41 years, 9 years earlier. Conclusions Multi-ancestry PRS approaches enhance lung adenocarcinoma risk stratification in East Asian never-smokers, with consistent external validation, suggesting future clinical utility.

Despite European colorectal liver metastases (CRLM) transplant trials having been performed entirely using deceased donor liver transplant (DDLT), there has been speculation about whether similar results can be achieved in the United States, given the lack of MELD score prioritization. A formal transplant protocol for unresectable CRLM was implemented in our center at the start of 2022 with the plan to utilize DDLT. Between 2022 and 2025, 86 patients with CRLM were referred to our program, and 19 patients underwent a formal transplant evaluation. Median time from listing until transplant was 31 days (range 3–115 d). Median allocation MELD was 8 (range 6–15). The donor livers were from DCDs, older donors, or National Share livers, which had been declined by all other centers. All but 2 of the cases underwent advanced perfusion. There were no cases of primary nonfunction, postreperfusion syndrome, acute kidney injury, hepatic artery thrombosis, or ischemic cholangiopathy. The length of stay for the median posttransplant was 5 days. Graft and patient survival were 100% at the time of maximal follow-up. Recurrence developed in 3 recipients (25.0%), 1 of whom was treated and has no evidence of disease. These findings support the viability of DDLT for CRLM using advanced perfusion, contingent on strict protocol adherence. Center-specific waitlist dynamics must inform the choice between DDLT and living donor liver transplant (LDLT).

Abstract Error analysis is a critical step in evaluating and improving clinical concept extraction models, the most common clinical natural language processing (NLP) task. Unlike corpus annotation, which follows standardized protocols for creating gold-standard datasets, error analysis requires nuanced judgment grounded in both clinical expertise and NLP knowledge. This is especially important given the heterogeneity of clinical text, where variations in documentation style, note structure, and terminology can substantially influence model behavior. Despite its importance, there is currently no standardized, user-level framework to support systematic error analysis in clinical concept extraction task. In this study, we developed and validated MedError, a machine-assisted, human-in-the-loop framework designed to standardize and enhance error analysis for clinical concept extraction tasks. We collected and manually curated a corpus of 1,187 unique errors from a total of 4,237 notes across three different distinct hospitals. The error categories were defined using our previously validated error taxonomy and included 480 false negatives and 707 false positives across 25 error types and 48 clinical concept categories. We evaluated the performance of three proprietary and three open-source large language models (LLMs) in automatically classifying these errors into 26 and 15 predefined categories. We further developed a machine-assisted framework, MedError, which integrates best practices in error analysis, LLM-assisted classification and reasoning, and a user-friendly interface to enable more efficient, reproducible, and context-aware error analysis. The framework supports both single-site and federated multisite error analysis, facilitating the effective translation of clinical NLP systems into real-world settings.