Yang Xie
Professor
Research Interests
Make your next move.
Prepare your application
Review drafts before you send them.
Explore related searches
Articles (13)
Background Current risk stratification for Ewing sarcoma relies heavily on clinical factors such as metastatic status, failing to capture histologic heterogeneity as a potential prognostic indicator. Although pathology reports contain rich biological data, this information remains locked in unstructured narrative text, limiting large-scale retrospective analyses. We aimed to validate the utility of Large Language Models (LLMs) for scalable data abstraction and to identify prognostic histologic features from a large multi-institutional cohort. Methods We conducted a retrospective cohort study using data from six Children’s Oncology Group (COG) clinical trials. We utilized an LLM-based pipeline (OpenAI o3) to extract structured variables—including immunohistochemical (IHC) markers and CD99 staining patterns—from digitized, Optical Character Recognition (OCR)-processed pathology reports. Extraction accuracy was validated against a human-annotated ground truth (n=200) and cross-validated against senior experts (n=48). We assessed the association between extracted features and Overall Survival (OS) using Kaplan-Meier analysis and multivariable Cox proportional hazards regression, adjusting for metastatic status. Findings We analyzed 931 diagnostic pathology reports spanning over 19-years. The LLM achieved a weighted average accuracy of 94% across 17 IHC markers; in a cross-validation subset, the LLM outperformed human annotators (weighted average accuracy over 15 IHC markers: LLM o3: 98.1%, a resident specialist 91.4%, and a senior expert 95.9%). Survival analysis identified Neuron-Specific Enolase (NSE) and S100 as significant prognostic biomarkers. After adjusting for metastatic status, NSE positivity was associated with significantly inferior survival (HR 2.15, 95% CI 1.15–4.02, p=0.016); this risk was most pronounced in patients with non-metastatic disease (HR 5.64, p=0.0055). Conversely, S100 positivity was associated with improved survival (HR 0.58, 95% CI 0.34–1.00, p=0.046). Interpretation LLM-assisted extraction of pathology variables is highly accurate and scalable, capable of unlocking “dark data” from historical clinical trials. We identified NSE as a potent risk factor and S100 as a protective marker in Ewing sarcoma, particularly in localized disease. These findings suggest that AI-derived histologic data can refine risk stratification and, if validated, warrant inclusion in future prospective trials. Research in Context Evidence before this study We searched PubMed and Google Scholar for articles published up to December 2025, using terms such as “Ewing sarcoma”, “prognosis”, “risk stratification”, “large language models”, “natural language processing”, and “pathology report extraction”. Current risk stratification for Ewing sarcoma relies predominantly on clinical variables, specifically the presence of metastatic disease and tumor size. While histologic heterogeneity is well-documented, it is rarely incorporated into risk models because extracting structured data from narrative pathology reports is labor-intensive. Small, single-institution studies have suggested potential prognostic roles for markers like Neuron-Specific Enolase (NSE) or S100, but results have been inconsistent and limited by small sample sizes. Furthermore, while Large Language Models (LLMs) have demonstrated potential for extracting clinical data, their application to “rescue” data from legacy, multi-institutional clinical trial documents (scanned PDF images) for rare disease biomarker discovery has not been validated at scale. Added value of this study To our knowledge, this is the largest study to utilize AI to extract histologic data from Ewing sarcoma pathology reports, aggregating 931 patients from six distinct Children’s Oncology Group (COG) clinical trials spanning over 19 years. We validated an LLM-based pipeline that converted noisy, optically character-recognized text from diverse institutions into structured data with 98.1% accuracy,
Year:
2026
Supplementary Figure S1 shows validation of key genetic features in NCI-237 and metabolic characterization of thyroid cancer cell lines. Supplementary Figure S2 shows growth curves and depletion/enrichment plots for pooled CRISPR-Cas9 screens. Supplementary Figure S3 shows validation of complex I impairment conferring sensitivity to LDH inhibition. Supplementary Figure S4 shows metabolic profiling of LDH inhibition in NCI-237 cells. Supplementary Figure S5 shows the validation of NADH-oxidizing enzymes and that NAD+ regeneration rescues LDH inhibition. Supplementary Figure S6 shows that LDH inhibition impairs NCI-237-R PDX proliferation.
Year:
2025
Identifying how drug molecular structures interact with specific gene targets remains a key challenge in cancer precision medicine. Although large-scale in vitro drug screening datasets are available, most computational models fail to elucidate the molecular mechanisms underlying drug-gene interactions or inform mechanisms of action (MOA). Here, we present MIDI (Mechanism-Interpretable Drug-Gene Interaction), a novel, attention-guided, mechanism-aware deep learning framework designed to address this gap. MIDI is trained on 80% of cell lines from the Cancer Cell Line Encyclopedia (CCLE) and evaluated on the remaining 20%, demonstrating improved drug response prediction performance over existing benchmarks. Leveraging chemical structure information, MIDI incorporates a Graphformer-based self-attention mechanism to capture atomic-level interactions between small molecules and gene features. A cross-attention module further enables the model to map drug structures to candidate gene targets, supporting interpretable mechanistic inference. To simulate a realistic in silico drug screening scenario, we evaluated MIDI on 52 previously unseen drugs from the Genomics of Drug Sensitivity in Cancer (GDSC) dataset. MIDI consistently ranked true gene targets among the top candidates across thousands of possibilities. These results highlight the feasibility and utility of in silico drug screening with MIDI, enabling both predictive modeling and mechanistic exploration. By uncovering interpretable drug-gene relationships, MIDI provides a scalable, data-driven framework to accelerate therapeutic discovery and target prioritization in preclinical research.
Year:
2025
Lung cancer, the leading cause of cancer mortality, exhibits diverse histologic subtypes and genetic complexities. Numerous preclinical mouse models have been developed to study lung cancer, but data from these models are disparate, siloed, and difficult to compare in a centralized fashion. In this study, we established the Lung Cancer Autochthonous Model Gene Expression Database (LCAMGDB), an extensive repository of 1,354 samples from 77 transcriptomic datasets covering 974 samples from genetically engineered mouse models (GEMM), 368 samples from carcinogen-induced models, and 12 samples from a spontaneous model. Meticulous curation and collaboration with data depositors produced a robust and comprehensive database, enhancing the fidelity of the genetic landscape it depicts. The LCAMGDB aligned 859 tumors from GEMMs with human lung cancer mutations, enabling comparative analysis and revealing a pressing need to broaden the diversity of genetic aberrations modeled in the GEMMs. To accompany this resource, a web application was developed that offers researchers intuitive tools for in-depth gene expression analysis. With standardized reprocessing of gene expression data, the LCAMGDB serves as a powerful platform for cross-study comparison and lays the groundwork for future research, aiming to bridge the gap between mouse models and human lung cancer for improved translational relevance. Significance: The Lung Cancer Autochthonous Model Gene Expression Database (LCAMGDB) provides a comprehensive and accessible resource for the research community to investigate lung cancer biology in mouse models.
Year:
2025
A metabolic hallmark of cancer identified by Warburg is the increased consumption of glucose and secretion of lactate, even in the presence of oxygen. Although many tumors exhibit increased glycolytic activity, most forms of cancer rely on mitochondrial respiration for tumor growth. We report here that Hürthle cell carcinoma of the thyroid (HTC) models harboring mitochondrial DNA-encoded defects in complex I of the mitochondrial electron transport chain exhibit impaired respiration and alterations in glucose metabolism. CRISPR–Cas9 pooled screening identified glycolytic enzymes as selectively essential in complex I–mutant HTC cells. We demonstrate in cultured cells and a patient-derived xenograft model that small-molecule inhibitors of lactate dehydrogenase selectively induce an ATP crisis and cell death in HTC. This work demonstrates that complex I loss exposes fermentation as a therapeutic target in HTC and has implications for other tumors bearing mutations that irreversibly damage mitochondrial respiration. Significance: HTC is enriched in somatic mtDNA mutations predicted to affect complex I of the electron transport chain (ETC). We demonstrate that these mutations impair respiration and induce a therapeutically tractable reliance on aerobic fermentation for cell survival. This work provides a rationale for targeting fermentation in cancers harboring irreversible genetically encoded ETC defects. See related article by Gopal et al., p. 1904. This article is highlighted in the In This Issue feature, p. 1749
Year:
2025
Lineage plasticity has long been documented in both small cell lung cancer (SCLC) and neuroblastoma (NBL), two clinically distinct neuroendocrine (NE) cancers. In this study, we quantified the NE features of cancer as NE scores and performed a systematic comparison of SCLC and NBL. We found NBL and SCLC cell lines have highly similar molecular profiles and shared therapeutic sensitivity. In addition, NE heterogeneity was observed at both the inter- and intra-cell line levels. Surprisingly, we did not find a significant association between NE scores and overall survival in SCLC or NBL. We described many shared and unique NE score-associated features between SCLC and NBL, including dysregulation of Myc oncogenes, alterations in protein expression, metabolism, drug resistance, and selective gene dependencies. Implications: Our work establishes a reference for molecular changes and vulnerabilities associated with NE to non-NE transdifferentiation through mutual validation of SCLC and NBL samples.
Year:
2024
A metabolic hallmark of cancer identified by Warburg is the increased consumption of glucose and secretion of lactate, even in the presence of oxygen. Although many tumors exhibit increased glycolytic activity, most forms of cancer rely on mitochondrial respiration for tumor growth. We report here that Hurthle cell carcinoma of the thyroid (HTC) models harboring mitochondrial DNA-encoded defects in complex I of the mitochondrial electron transport chain exhibit impaired respiration and alterations in glucose metabolism. CRISPR-Cas9 pooled screening identified glycolytic enzymes as selectively essential in complex I-mutant HTC cells. We demonstrate in cultured cells and a PDX model that small molecule inhibitors of lactate dehydrogenase selectively induce an ATP crisis and cell death in HTC. This work demonstrates that complex I loss exposes fermentation as a therapeutic target in HTC and has implications for other tumors bearing mutations that irreversibly damage mitochondrial respiration.
Year:
2024
Supplementary Figure S1 shows validation of key genetic features in NCI-237 and metabolic characterization of thyroid cancer cell lines. Supplementary Figure S2 shows growth curves and depletion/enrichment plots for pooled CRISPR-Cas9 screens. Supplementary Figure S3 shows validation of complex I impairment conferring sensitivity to LDH inhibition. Supplementary Figure S4 shows metabolic profiling of LDH inhibition in NCI-237 cells. Supplementary Figure S5 shows the validation of NADH-oxidizing enzymes and that NAD+ regeneration rescues LDH inhibition. Supplementary Figure S6 shows that LDH inhibition impairs NCI-237-R PDX proliferation.
Year:
2024
Supplementary Figure S1 shows validation of key genetic features in NCI-237 and metabolic characterization of thyroid cancer cell lines. Supplementary Figure S2 shows growth curves and depletion/enrichment plots for pooled CRISPR-Cas9 screens. Supplementary Figure S3 shows validation of complex I impairment conferring sensitivity to LDH inhibition. Supplementary Figure S4 shows metabolic profiling of LDH inhibition in NCI-237 cells. Supplementary Figure S5 shows the validation of NADH-oxidizing enzymes and that NAD+ regeneration rescues LDH inhibition. Supplementary Figure S6 shows that LDH inhibition impairs NCI-237-R PDX proliferation.
Year:
2024
Patient-derived xenografts (PDX) model human intra- and intertumoral heterogeneity in the context of the intact tissue of immunocompromised mice. Histologic imaging via hematoxylin and eosin (H&E) staining is routinely performed on PDX samples, which could be harnessed for computational analysis. Prior studies of large clinical H&E image repositories have shown that deep learning analysis can identify intercellular and morphologic signals correlated with disease phenotype and therapeutic response. In this study, we developed an extensive, pan-cancer repository of >1,000 PDX and paired parental tumor H&E images. These images, curated from the PDX Development and Trial Centers Research Network Consortium, had a range of associated genomic and transcriptomic data, clinical metadata, pathologic assessments of cell composition, and, in several cases, detailed pathologic annotations of neoplastic, stromal, and necrotic regions. The amenability of these images to deep learning was highlighted through three applications: (i) development of a classifier for neoplastic, stromal, and necrotic regions; (ii) development of a predictor of xenograft-transplant lymphoproliferative disorder; and (iii) application of a published predictor of microsatellite instability. Together, this PDX Development and Trial Centers Research Network image repository provides a valuable resource for controlled digital pathology analysis, both for the evaluation of technical issues and for the development of computational image–based methods that make clinical predictions based on PDX treatment studies. Significance: A pan-cancer repository of >1,000 patient-derived xenograft hematoxylin and eosin–stained images will facilitate cancer biology investigations through histopathologic analysis and contributes important model system data that expand existing human histology repositories.
Year:
2024
Year:
2024
Year:
2024
Year:
2023
Collaborators (56)
Matthew H Bailey
Assistant Professor
Brigham Young University
Kekoa Taparra
Assistant Professor
University of California, Los Angeles
Michael A. Davies
-
John Minna
Professor and Direcdtor
University of Texas Southwestern Medical Center
Vassilis Aidinis
Biomedical Sciences Research Center Alexander Fleming
Barbara Grüner
University Hospital Essen
Sherri Davies
Senior Scientist
Washington University in St. Louis
Phuoc T. Tran
University of Maryland
Pawel Mazur
Associate Professor
The University of Texas MD Anderson Cancer Center
Roger A. Moorehead
University of Guelph
Ralph J. DeBerardinis
University of Texas Southwestern Medical Center
Charles Rudin
Memorial Sloan Kettering Cancer Center
Patrick Leavey
Southwestern Medical Center
J. Edward van Veen
Assistant Adjunct Professor
University of California, Los Angeles
Xiaowei Zhan
University of Texas Southwestern Medical Center
George Acquaah-Mensah
Professor of Pharmaceutical Sciences
Massachusetts College of Pharmacy and Health Sciences
Peter Robinson
Professor for Medical Computer Science
The Jackson Laboratory for Genomic Medicine
Jeffrey Chuang
Jackson Laboratory
Yonathan Lissanu
The University of Texas MD Anderson Cancer Center
Trudy G. Oliver
Duke University
Xiaowei Xu
Professor
University of Pennsylvania
YING GAO
University of Texas Southwestern Medical Center
Shidan Wang
-
David MacPherson
Professor
Fred Hutch Cancer Center
Giorgio Ramadori
Scientific Collaborator
University of Geneva
Rosanna Martinelli
Professor of Molecular Biology
Università degli Studi di Napoli Federico II
Junya Fujimoto
Professor and Chair
Hiroshima University Hospital
Alana Welm
Chair
Huntsman Cancer Institute
David G. McFadden
University of Texas Southwestern Medical Center
Min Gyu Lee
Professor
The University of Texas MD Anderson Cancer Center
Nicholas Mitsiades
Professor
University of California, Davis
Bryan Welm
Associate Professor
Huntsman Cancer Institute
Meenhard Herlyn
Professor, Director
The Wistar Institute
Lin Xu
Assistant Professor (tenure track)
University of Texas Southwestern Medical School
Vicky Li
National Institutes of Health
Shunqiang Li
Associate Professor
Washington University in St. Louis
Ryan C. Fields
University of Rochester
Luis G Carvajal‐Carmona
Professor
University of California, Davis
Noelle Williams
Research Professor
University of Texas Southwestern Medical Center
Funda Meric-Bernstam
The University of Texas MD Anderson Cancer Center
Robert Mashl
Washington University in St. Louis
Chong-xian Pan
Professor
university of california davis medical center
Carol Bult
Professor and Knowlton Family Chair
Jackson Laboratory
Salma Kaochar
Assistant Professor
Baylor College of Medicine
Li Ding
Associate Professor
Washington University in Saint Louis School of Medicine
Andrew Y. Koh
Professor and Division Chief
University of Texas Southwestern Medical Center
Shyam Biswal
Professor
Johns Hopkins University
Michael Lewis
Baylor College of Medicine
Arun Pandiri
Staff Scientist
National Institute of Environmental Health Sciences
CHARLES POWELL
Division Chief/Professor of Medicine
Icahn School of Medicine at Mount Sinai
Maria Gabriela Raso
Assistant Professor
The University of Texas MD Anderson Cancer Center
Soner Koc
GlaxoSmithKline Research and Development
Jennifer Beane-Ebel
Assistant Professor
Boston University School of Medicine
Guanghua Xiao
-
Jingwei Huang
University of Texas Southwestern Medical Center
Prashant Mishra
Associate Professor
University of Texas Southwestern Medical Center
Make your next move.
Prepare your application
Review drafts before you send them.