DSDavid J. H. Shih

David J. H. Shih

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


David J. H. Shih has a total of 992 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Top Cited Papers
    Open Access

    Multidisciplinary blinded randomized expert evaluation of large language models for clinical diagnosis and management

    by Peikai Chen, Jifu Cai, Jiaying Zhou, Shaoxi Chen, Chenguang Xu, Lihua Yuan, Xiaoying Dai, Xiaowei Chen, Yanzhe Wei, Xia Li, et al.

    Background Direct clinical uses of large language models (LLMs) remain controversial, partly because of the lack of methodological rigor in assessing their risks and benefits in medicine. Methods We developed Medieval, a multidisciplinary, randomized, and blinded expert evaluation framework. A ten-point Dreyfus-based scoring scale linked to career stages of human physicians was designed to reflect response qualities. Seven advanced LLMs or their distilled versions that were released within a short time-frame ( ≤ 45 days) in early 2025 were tested. Incidence of fabricated medical facts were documented. Linear mixed-effects models and variance-stabilizing Bayesian generalized linear mixed models were employed to perform statistical analyses. Results We first develop a high-quality question bank comprising 685 real and simulated clinical cases across 13 specialties. An expert panel of 27 clinicians (average years of services: 25.9) evaluated the 4795 model responses. We show that these LLM ratings (n = 9856) have excellent reliability (intraclass correlation coefficients \( > \)0.9). Among the seven LLMs tested, Gemini 2.0 Flash achieved the highest raw scores. However, after adjusting for confounders, DeepSeek-R1 was the top-performing model with a mean score of 6.36 (95% confidence interval 6.03 − 6.69), a performance level equivalent to an early-career physician. Despite these strengths, 3–19% LLM responses were rated as incompetent and 40 instances of LLM hallucination were also identified. Conclusions Our study shows that in spite of LLMs’ substantial potentials in medicine, their unguarded clinical application could present serious risks, which must be continuously monitored by human expert panels. The evaluation framework developed and validated in this study will facilitate such efforts.


  • Top Cited Papers
    Open Access

    Intratumoral bacteria are immunosuppressive and promote immunotherapy resistance in head and neck squamous cell carcinoma

    by Natalie L. Silver, Jin Dai, Travis D. Kerr, Jessica Altemus, Rekha Garg, Hannah Simmons, Tyler Alban, Laura Noel-Romas, Vladimir Makarov, David J. H. Shih, et al.

    Despite the promise of immune checkpoint blockade (ICB) in head and neck squamous cell carcinoma (HNSCC), mediators of response are poorly understood. To address this, here we analyzed oropharyngeal HNSCCs treated with neoadjuvant durvalumab (anti-PDL1) alone or in combination with tremelimumab (anti-CTLA4) from the CIAO clinical trial ( NCT03144778 ). We found that only the total abundance of intratumoral bacteria predicted ICB response, which was validated in multiple independent cohorts. High intratumoral bacteria abundance was associated with an immunosuppressive tumor microenvironment, characterized by an accumulation of neutrophils coupled with depletion of T cells and other adaptive immune cells. Experimental elevation or reduction in intratumoral bacteria abundance in orthotopic models of HNSCC in female mice recapitulated immunological associations observed in participant tumors. Increasing intratumoral bacteria abundance was sufficient to induce resistance to anti-PDL1 ICB, irrespective of bacterial species tested. Together, these findings demonstrate that high intratumoral bacteria abundance is a key suppressor of antitumor immunity and promotes immunotherapy resistance.


  • Open Access

    enLLASD: An Ensemble Deep Learning Framework to Automate Derivation of Lower Limb Alignments for Skeletal Dysplasia

    by Peikai Chen, Xinlin Zhou, Haihua Cai, David J. H. Shih, Janus S. H. Wong, Yong Hu, Michael Kai-Tsun To

    Skeletal dysplasia (SD) is a group of rare, congenital skeletal disorders that affect millions worldwide. Patients often present with moderate to severe limb deformities, which need to be constantly monitored radiographically, in a process called alignment assessments. Currently, alignments are usually determined manually by physicians, with significant inter-rater variability and low efficiencies. Recent efforts to automate this often used individual deep-learning models to detect bone contours or landmarks (e.g. joints). Due to data scarcity, case heterogeneity, low bone mass, existing methods risk model overfitting. We propose enLLASD, an ensemble learning framework that integrates key-point detection and bone segmentation results from multiple member models, to enhance robustness. Predictions are aggregated using averaging, majority voting, and logistic stacking, while false positives in segmentation are suppressed through consistency with key-point predictions. A spline-based resampling method is employed to fit the medial axes of the femur and tibia, enabling the computation of key alignment angles. We validated our framework on a dataset of 1,416 full-length radiographs from both SD and non-SD individuals. Experimental results demonstrate that logistic stacking (IOU [intersection over union] for front femur: 0.9564, front tibia: 0.9468) outperforms individual models in segmentation accuracy and alignment angle estimation, particularly in cases with severe deformities. Our work highlights the potential of ensemble deep learning in automatic orthopedic radiography for SD.

  • Open Access

    Tumor-associated macrophage-derived exosomes modulate the immunotherapeutic sensitivity of SHH-medulloblastoma by targeting m6A-modified FOXD1

    by Yantao Liu, Yu Peng, Chen Song, Zongran Liu, Xiaolong Yang, Shuqing Bian, Xiaolin Xiao, Haishuang Li, Jing Wang, Ziwen Sun, et al.

    Background: Medulloblastoma (MB) is the most common pediatric malignant brain tumor. Infiltration of tumor-associated macrophages (TAMs) and m6A modification of RNA are correlated with poor prognosis and tumor progression in the Sonic Hedgehog (SHH) subtype (SHH-MB). However, the relationship between TAMs infiltration in SHH-MB and m6A modification status during tumor progression remains unclear. Methods: Expression of m6A modification-related proteins was assessed in 40 cases of SHH-MB. Genes affected by TAM-derived exosomes were identified with methylated RNA immunoprecipitation sequencing. Mechanisms of m6A modification of FOXD1 were evaluated and combinatorial treatment with AAV2/9-shFOXD1 and PD-1 inhibitors was investigated in the NeuroD2:SmoA1 mouse model. Results: TAMs infiltration led to decreased METTL14 expression, which was mediated by TAM-derived exosomes containing METTL14-specific microRNAs. In turn, this led to lower levels of m6A modifications. Through a screen, FOXD1 was identified as a critical downstream target of TAM-derived exosomes, and its expression level was correlated with poor prognosis in SHH-MBs. Importantly, knockdown of FOXD1 in SHH-MB cells significantly promoted the release of chemokines CXCL10/11, resulting in CD8+ T cell recruitment. Furthermore, treatment with AAV2/9-shFOXD1 significantly enhanced the antitumor effect of the PD-1 inhibitor in transgenic SHH-MB mice. Conclusions: Our study revealed for the first time that TAM-derived exosomes modulate m6A levels in SHH-MB, which promotes tumor progression via FOXD1. We identified FOXD1 as a novel therapeutic target whose inhibition sensitizes SHH-MB to immune checkpoint blockade.


  • Open Access

    The IL6/JAK/STAT3 signaling axis is a therapeutic vulnerability in SMARCB1-deficient bladder cancer

    by Chandra Sekhar Amara, Karthik Reddy Kami Reddy, Yang Yuntao, Yuen San Chan, Danthasinghe Waduge Badrajee Piyarathna, Lacey Elizabeth Dobrolecki, David J. H. Shih, Zhongcheng Shi, Jun Xu, Shixia Huang, et al.

    SMARCB1 loss has long been observed in many solid tumors. However, there is a need to elucidate targetable pathways driving growth and metastasis in SMARCB1-deficient tumors. Here, we demonstrate that SMARCB1 deficiency, defined as genomic SMARCB1 copy number loss associated with reduced mRNA, drives disease progression in patients with bladder cancer by engaging STAT3. SMARCB1 loss increases the chromatin accessibility of the STAT3 locus in vitro. Orthotopically implanted SMARCB1 knockout (KO) cell lines exhibit increased tumor growth and metastasis. SMARCB1-deficient tumors show an increased IL6/JAK/STAT3 signaling axis in in vivo models and patients. Furthermore, a pSTAT3 selective inhibitor, TTI-101, reduces tumor growth in SMARCB1 KO orthotopic cell line-derived xenografts and a SMARCB1-deficient patient derived xenograft model. We have identified a gene signature generated from SMARCB1 KO tumors that predicts SMARCB1 deficiency in patients. Overall, these findings support the clinical evaluation of STAT3 inhibitors for the treatment of SMARCB1-deficient bladder cancer.


  • Open Access

    Biomolecular Condensation: A New Phase in Cancer Research

    by Anupam K. Chakravarty, Daniel J. McGrail, Thomas M. Lozanoski, Brandon S. Dunn, David J.H. Shih, Kara M. Cirillo, Sueda H. Cetinkaya, Wenjin Jim Zheng, Gordon B. Mills, S. Stephen Yi, et al.

    Multicellularity was a watershed development in evolution. However, it also meant that individual cells could escape regulatory mechanisms that restrict proliferation at a severe cost to the organism: cancer. From the standpoint of cellular organization, evolutionary complexity scales to organize different molecules within the intracellular milieu. The recent realization that many biomolecules can “phase-separate” into membraneless organelles, reorganizing cellular biochemistry in space and time, has led to an explosion of research activity in this area. In this review, we explore mechanistic connections between phase separation and cancer-associated processes and emerging examples of how these become deranged in malignancy. Significance: One of the fundamental functions of phase separation is to rapidly and dynamically respond to environmental perturbations. Importantly, these changes often lead to alterations in cancer-relevant pathways and processes. This review covers recent advances in the field, including emerging principles and mechanisms of phase separation in cancer.


  • Open Access

    CNGPLD: case–control copy-number analysis using Gaussian process latent difference

    by David J H Shih, Ruoxing Li, Peter Müller, W Jim Zheng, Kim-Anh Do, Shiaw-Yih Lin, Scott L Carter

    Cross-sectional analyses of primary cancer genomes have identified regions of recurrent somatic copy-number alteration, many of which result from positive selection during cancer formation and contain driver genes. However, no effective approach exists for identifying genomic loci under significantly different degrees of selection in cancers of different subtypes, anatomic sites or disease stages. CNGPLD is a new tool for performing case–control somatic copy-number analysis that facilitates the discovery of differentially amplified or deleted copy-number aberrations in a case group of cancer compared with a control group of cancer. This tool uses a Gaussian process statistical framework in order to account for the covariance structure of copy-number data along genomic coordinates and to control the false discovery rate at the region level. CNGPLD is freely available at https://bitbucket.org/djhshih/cngpld as an R package. Supplementary data are available at Bioinformatics online.


  • Open Access

    Role of DNA repair defects in predicting immunotherapy response

    by Jing Zhang, David J. H. Shih, Shiaw-Yih Lin

    Defect in DNA damage response (DDR) is a common feature of cancer cells, which regulates tumor growth and therapeutic response. Recently, the approval of immune checkpoint blockade (ICB) for tumors with defective mismatch repair has paved the way for investigating the role of other DDR defects in sensitizing cancer to ICB therapy. Despite great progress in understanding DDR pathways, the mechanisms that link DDR defects and ICB response remain incompletely understood. Further, the clinical activity of ICB in patients with DDR defective tumors has not been well described. Here, we discuss recent studies demonstrating that biomarkers in DDR pathways may serve as potential predictors to guide the selection of patients for ICB therapy. A better understanding of the relationship between deficiency in DDR and response to ICB would facilitate efforts in optimizing the efficacy of immunotherapy.


  • Open Access

    Modeling germline mutations in pineoblastoma uncovers lysosome disruption-based therapy

    by Philip E. D. Chung, Deena M. A. Gendoo, Ronak Ghanbari-Azarnier, Jeff C. Liu, Zhe Jiang, Jennifer Tsui, Dong-Yu Wang, Xiao Xiao, Bryan Li, Adrian Dubuc, et al.

    Pineoblastoma is a rare pediatric cancer induced by germline mutations in the tumor suppressors RB1 or DICER1. Presence of leptomeningeal metastases is indicative of poor prognosis. Here we report that inactivation of Rb plus p53 via a WAP-Cre transgene, commonly used to target the mammary gland during pregnancy, induces metastatic pineoblastoma resembling the human disease with 100% penetrance. A stabilizing mutation rather than deletion of p53 accelerates metastatic dissemination. Deletion of Dicer1 plus p53 via WAP-Cre also predisposes to pineoblastoma, albeit with lower penetrance. In silico analysis predicts tricyclic antidepressants such as nortriptyline as potential therapeutics for both pineoblastoma models. Nortriptyline disrupts the lysosome, leading to accumulation of non-functional autophagosome, cathepsin B release and pineoblastoma cell death. Nortriptyline further synergizes with the antineoplastic drug gemcitabine to effectively suppress pineoblastoma in our preclinical models, offering new modality for this lethal childhood malignancy.


  • Open Access

    Cellular origins and genetic landscape of cutaneous gamma delta T cell lymphomas

    by Jay Daniels, Peter G. Doukas, Maria E. Martinez Escala, Kimberly G. Ringbloom, David J. H. Shih, Jingyi Yang, Kyle Tegtmeyer, Joonhee Park, Jane J. Thomas, Mehmet E. Selli, et al.

    Primary cutaneous γδ T cell lymphomas (PCGDTLs) represent a heterogeneous group of uncommon but aggressive cancers. Herein, we perform genome-wide DNA, RNA, and T cell receptor (TCR) sequencing on 29 cutaneous γδ lymphomas. We find that PCGDTLs are not uniformly derived from Vδ2 cells. Instead, the cell-of-origin depends on the tissue compartment from which the lymphomas are derived. Lymphomas arising from the outer layer of skin are derived from Vδ1 cells, the predominant γδ cell in the epidermis and dermis. In contrast, panniculitic lymphomas arise from Vδ2 cells, the predominant γδ T cell in the fat. We also show that TCR chain usage is non-random, suggesting common antigens for Vδ1 and Vδ2 lymphomas respectively. In addition, Vδ1 and Vδ2 PCGDTLs harbor similar genomic landscapes with potentially targetable oncogenic mutations in the JAK/STAT, MAPK, MYC, and chromatin modification pathways. Collectively, these findings suggest a paradigm for classifying, staging, and treating these diseases.