publications

2026

1. bioRxiv

   PromptBio-Bench: Benchmarking LLM-based Bioinformatics Agents for End-to-End Data Analysis

   Wenbin Guo, Minzhe Zhang, Bowei Han, and 7 more authors

   bioRxiv, 2026

   Abs DOI PDF

   Large language model (LLM)-based agents hold transformative potential for automating bioinformatics workflows; however, systematic evaluations of their capabilities remain limited, hindering a clear assessment of their readiness for real-world application. We introduce PromptBio-Bench, a comprehensive evaluation suite of 194 expert-curated tasks spanning bioinformatics and data science at varied difficulty levels, and an evaluation framework for structured file comparison and scoring against expert reference answers. Benchmarking three state-of-the-art agents revealed that Biomni and ToolsGenie achieved comparable performance, and accuracy declined markedly at higher difficulty levels across all agents. As foundation models and agent frameworks continue to evolve, PromptBio-Bench provides a valuable benchmark infrastructure for the community to systematically track the progress of agentic bioinformatics.

2025

1. bioRxiv

   Dual-LLM Adversarial Framework for Information Extraction from Research Literature

   Zhijing Li, Yunwen Yu, Wenhao Gu, and 5 more authors

   bioRxiv, 2025

   Abs DOI PDF

   Information Extraction (IE) is a fundamental task in Natural Language Processing (NLP) that aims to automatically identify relevant information from unstructured or semi-structured data. Information extraction from lengthy research literature, particularly in multi-omics studies, faces significant challenges due to their complex narratives and extensive context. To address this, we present a novel dual–LLM adversarial framework in which one large language model (LLM) performs the extraction and another provides iterative feedback to refine the results. This process systematically reduces errors, enhances consistency across heterogeneous data sources, and converges toward more accurate outputs. We evaluated our approach against manual and single-LLM extraction, using LLMs as evaluators. Experimental results show that our adversarial framework outperforms these baselines, highlighting its effectiveness for extracting structured information from lengthy scientific texts.

2. bioRxiv

   PromptBio: A Multi-Agent AI Platform for Bioinformatics Data Analysis

   Minzhe Zhang, Wenhao Gu, Bowei Han, and 16 more authors

   bioRxiv, 2025

   Abs DOI PDF

   PromptBio is a modular AI platform for scalable, reproducible, and user-adaptable bioinformatics analysis, powered by generative AI and natural language interaction. It supports three complementary modes of analysis designed to meet diverse research needs. PromptGenie is a multi-agent system that enables stepwise, human-in-the-loop workflows using prevalidated domain-standard tools. Within PromptGenie, specialized agents—including DataAgent, OmicsAgent, AnalysisAgent, and QAgent—collaborate to manage tasks such as data ingestion, pipeline execution, statistical analysis, and interactive summarization. DiscoverFlow provides integrated, automated workflows for large-scale multi-omics analysis, offering end-to-end execution and streamlined orchestration. ToolsGenie complements these modes by dynamically generating executable bioinformatics code for custom, user-defined analyses, enabling flexibility beyond standardized workflows. PromptGenie and DiscoverFlow leverage a suite of domain-specific tools, including Omics Tools for standardized omics pipelines, Analysis Tools for downstream statistical interpretation, and MLGenie for machine learning and multi-omics modeling. We present the design, capabilities, and validation of these components, highlight their integration into automated and customizable workflows, and discuss extensibility, monitoring, and compliance. PromptBio aims to democratize high-throughput bioinformatics through a large language model–powered, natural language understanding, workflow generation and agent orchestration.

3. Med

   Overcoming ovarian cancer resistance and evasion to CAR-T cell therapy by harnessing allogeneic CAR-NKT cells

   Yan-Ruide Li, Zhe Li, Yichen Zhu, and 38 more authors

   Med, 2025

   Abs DOI PDF

   Ovarian cancer remains one of the most lethal gynecologic malignancies, often recurring due to treatment resistance and immune evasion. Current chimeric antigen receptor-engineered T (CAR-T) cell therapies have shown limited success against ovarian cancer, largely due to tumor heterogeneity and an immunosuppressive tumor environment. This study explores an alternative approach using allogeneic CAR-engineered invariant natural killer T (AlloCAR-NKT) cells derived from blood stem cells. By applying a clinically guided culture method, the researchers generated AlloCAR-NKT cells with high efficiency, purity, and potent antitumor activity. These engineered cells not only directly target tumor cells but also remodel the tumor environment and display a favorable safety profile. The findings suggest that AlloCAR-NKT cells offer a safer, more effective, and readily available immunotherapy for patients with ovarian cancer.

2024

1. medRxiv

   Type-2 diabetes biomarker discovery and risk assessment through saliva DNA methylome

   Wenbin Guo, Marco Morselli, Kimberly C. Paul, and 3 more authors

   medRxiv, 2024

   Abs DOI PDF

   The rising prevalence of type 2 diabetes (T2D) motivates innovative strategies to deepen disease understanding and enhance diagnostic capabilities. This study measures diabetes-specific epigenetic signals in saliva, establishing saliva DNA methylome as a promising medium for T2D screening and study. By integrating comprehensive whole-genome bisulfite sequencing (WGBS) and high-depth targeted bisulfite sequencing (TBS), we developed a cost-efficient two-step approach to profiling DNA methylation at regions of interest. WGBS analysis confirmed T2D-specific methylation signatures in saliva, revealing their enrichment in immune and metabolic regulation pathways. TBS enabled accurate cell type deconvolution, revealing minimal differences in cellular composition between diabetic and non-diabetic samples, suggesting intrinsic molecular changes drive the observed methylation changes. Epigenome-wide association studies further identified significant CpG sites, notably in the ABCG1 region, with strong potential for T2D status prediction. These findings validate the saliva DNA methylome as a scalable, non-invasive resource for T2D biomarker discovery, advancing opportunities in T2D screening, risk assessment, and personalized medicine.

2. bioRxiv

   Systematic dissection of epigenetic age acceleration in normal breast tissue reveals its link to estrogen signaling and cancer risk

   Mary E. Sehl, Wenbin Guo, Collin Farrell, and 8 more authors

   bioRxiv, 2024

   Abs DOI PDF

   Breast aging encompasses intricate molecular and cellular changes that elevate cancer risk. Our study profiled DNA methylation and gene expression of 181 normal breast samples and systematically evaluated eight epigenetic clocks. We found that clocks trained using breast tissues demonstrate improved age prediction in normal breast tissue, and bias universally exists in epigenetic clocks, necessitating a proper definition of age acceleration. Cell composition analysis revealed significant age-related alterations and highlighted its distinct associations with age acceleration, including increased luminal epithelial and myoepithelial cells and reduced adipocytes and immune cells, connecting age acceleration to carcinogenesis from a cell compositional perspective. Additionally, CpG sites associated with age acceleration were enriched for estrogen receptor binding sites, providing a mechanistic link between estrogen exposure, accelerated aging, and cancer. These findings highlight the importance of cellular heterogeneity in epigenetic age estimates and the potential of age acceleration to guide for risk stratification and prevention strategies.

3. bioRxiv

   BSReadSim: a versatile and efficient simulator to generate realistic bisulfite sequencing reads

   Wenbin Guo, and Matteo Pellegrini

   bioRxiv, 2024

   Abs DOI PDF

   Realistic bisulfite sequencing simulators are crucial for advancing method development in computational epigenetics. However, existing tools often fall short due to oversimplified generative models that fail to capture the complexity of real data. We present BSReadSim, an efficient and versatile simulator that generatesrealistic bisulfite sequencing reads. BSReadSim excels in integrating reference genetic variants and methylation profiles, offering unmatched versatility across multiple sequencing technologies, including WGBS, RRBS, and TBS. By accurately modeling methylation patterns, sampling biases, sequencing errors, and leveraging optimized implementation, BSReadSim efficiently generates realistic synthetic datasets tailored to specific experimental needs while maintaining computational feasibility. By enhancing the realism and flexibility of bisulfite sequencing simulations, BSReadSim supports improved experiment design, method development, and benchmarking of computational tools, ultimately advancing the reliability and rigor of DNA methylation analysis tools.

4. Master’s Thesis

   Modeling Site-Site Dependency in DNA Methylation Sequencing data

   Wenbin Guo

   UCLA, 2024

   Abs Link PDF

   DNA methylation is a critical epigenetic mark that regulates gene expression and plays a crucial role in various biological processes. However, the complexity of DNA methylation data makes it challenging to develop accurate computational models. This thesis presents a comprehensive study of site-site dependency in DNA methylation sequencing data, aiming to develop a more accurate computational model for DNA methylation analysis.

5. PhD Thesis

   Computational method development and analysis for DNA methylome studies

   Wenbin Guo

   UCLA, 2024

   Abs Link PDF

   DNA methylation is a critical epigenetic mark that regulates gene expression and plays a crucial role in various biological processes. However, the complexity of DNA methylation data makes it challenging to develop accurate computational models. This thesis presents a comprehensive study of site-site dependency in DNA methylation sequencing data, aiming to develop a more accurate computational model for DNA methylation analysis.

6. Nature Biotech

   Generation of allogeneic CAR-NKT cells from hematopoietic stem and progenitor cells using a clinically guided culture method

   Yan-Ruide Li, Yang Zhou, Jiaji Yu, and 35 more authors

   Nature Biotechnology, 2024

   Abs DOI PDF

   Allogeneic chimeric antigen receptor-engineered natural killer T (CAR-NKT) cells have shown promising antitumor activity in preclinical models. However, the generation of these cells from hematopoietic stem and progenitor cells (HSPCs) is challenging due to their low frequency and the lack of clinically validated culture methods. We developed a clinically guided culture method to generate allogeneic CAR-NKT cells from HSPCs. We found that the culture method is effective in generating CAR-NKT cells with high efficiency, purity, and antitumor activity. These findings pave the way for the clinical application of allogeneic CAR-NKT cells in ovarian cancer therapy.

2023

1. Nature Comm

   Unlocking the potential of allogeneic V\delta2 T cells for ovarian cancer therapy through CD16 biomarker selection and CAR/IL-15 engineering

   Derek Lee, Zachary Spencer Dunn, Wenbin Guo, and 14 more authors

   Nature Communications, 2023

   Abs DOI PDF

   Ovarian cancer is a highly aggressive disease with limited treatment options. Current therapies often fail due to tumor heterogeneity and immune evasion. We developed a novel approach using allogeneic V\delta2 T cells engineered with CD16 and CAR/IL-15, targeting ovarian cancer cells with high efficacy and safety. Our study demonstrates that CD16 selection is crucial for CAR/IL-15 engineering and that the engineered cells can effectively target ovarian cancer cells and remodel the tumor microenvironment. These findings pave the way for personalized immunotherapy strategies in ovarian cancer.

2018

1. Cancer Discov

   An Empirical Approach Leveraging Tumorgrafts to Dissect the Tumor Microenvironment in Renal Cell Carcinoma Identifies Missing Link to Prognostic Inflammatory Factors

   Tao Wang, Rong Lu, Payal Kapur, and 28 more authors

   Cancer Discovery, 2018

   Abs DOI PDF

   Renal cell carcinoma (RCC) is a common kidney cancer with high mortality. The tumor microenvironment (TME) plays a crucial role in RCC progression and prognosis. However, the TME is complex and heterogeneous, making it challenging to dissect its components. We developed a novel approach using tumorgrafts to dissect the TME in RCC. We found that the TME is enriched in inflammatory factors, which are associated with poor prognosis. These findings pave the way for personalized treatment strategies in RCC.