About

Jeremy Lyle Warner is a Professor of Medicine and Professor of Biostatistics at Brown University. He serves as the Associate Director of Data Science at the Legorreta Cancer Center and is an attending physician at the Lifespan Cancer Institute. Board certified in Medical Oncology, Hematology, and Clinical Informatics, Dr. Warner's clinical focus is malignant hematology. His primary research goal is to make sense of the structured and unstructured data present in electronic health records (EHRs) and clinical knowledge bases, with the aim of directly improving clinical care for patients with cancer. As a practicing clinician with a background in engineering, medicine, and biomedical informatics, he has contributed to the establishment of clinical genomics standards and large-scale initiatives around knowledge curation and management. His work involves the extraction, standardization, and presentation of clinical data to various stakeholders, including clinicians, researchers, and patients. Dr. Warner is also the Editor-in-Chief of JCO Clinical Cancer Informatics, the Chief Technology Officer of HemOnc.org, and the founding Director of the Center for Clinical Cancer Informatics and Data Science at Brown University. He co-founded the COVID-19 and Cancer Consortium (CCC19) and directs its Research Coordinating Center.

Research topics

• Medicine
• Internal medicine
• Oncology
• Statistics
• Pathology
• Genetics
• Biology
• Computational biology
• Bioinformatics
• Intensive care medicine
• Environmental health

Selected publications

• Inclusion of people living with HIV in Food and Drug Administration (FDA) oncology pivotal registration trials from 2020 to 2024.

  Journal of Clinical Oncology · 2025-05-28 · 2 citations

  article

  1517 Background: People living with HIV (PLWH) have an increased risk of developing cancers compared to the population without HIV, with cancer being the leading cause of death for this population in high-income countries. Previous research by the ASCO-led HIV working group found only 11% of clinical trials supporting FDA cancer therapy approvals from 2010-2014 allowed for inclusion of PLWH, leading to clinical uncertainty in the efficacy-safety profile of new cancer treatments in this group. To address this gap, ASCO (2017) and FDA (2020) issued guidelines encouraging inclusion of PLWH in cancer clinical trials. To explore their impact, we examined inclusion of PLWH in pivotal cancer trials post-guidelines release. Methods: We reviewed all new FDA-approved indications of the past five years (Jan/2020-Nov/2024). For each new approval, two authors independently assessed the inclusion/exclusion criteria outlined in the primary protocol of each pivotal trial. We analyzed data on the FDA label, cancer type, therapeutic modality, inclusion/exclusion criteria, sponsor, and the protocol (version 1) publication date. Results: We identified 244 new therapy indications, based on supporting data from 259 pivotal clinical trials. 27% of trials permitted inclusion of PLWH. Pivotal trials for hematological cancers, compared to solid cancers, were significantly more likely to exclude PLWH [unadjusted Odds Ratio (OR) 3.15, 95% confidence interval (CI): 1.51-6.56; p=0.0012]. Pivotal trials of immunomodulatory agents were significantly more likely (OR 3.87, 95% CI: 1.91–7.83; p<0.0001) to exclude PLWH compared to other cancer therapies. The inclusion rate was 10.3% for AIDS-defining cancers and 29.8% for non-AIDS-defining cancers. Trials funded only by industry were significantly more likely (OR 2.80, 95% CI: 1.36-5.77; p=0.0078) to exclude PLWH, compared to non-industry funded trials. Inclusion rate of PLWH was higher in protocols published after 2020 (39.1%) compared to those before (26.3%). Conclusions: Our analysis indicates an improvement in the inclusion of PLWH in oncology pivotal trials following ASCO and FDA guidance. However, nearly three out of four pivotal cancer trials continue to exclude PLWH. This highlights an unmet need, resulting in uncertainty for healthcare professionals regarding the safety and clinical utility of new cancer treatments in PLWH. Additional strategies must be considered to address this disparity. Rates of PLWH inclusion in oncology pivotal trials. Include PLWH (%) Total (n=259) 27.4 Solid Cancers (185) 33 Haematological malignancies (74) 13.5 Immunomodulatory agents (89) 12.4 Other agents*(170) 35.3 Industry-funded (223) 24.2 Non-industry funded (36) 47.2 AIDS-Defining Cancers (30) 10 Non-AIDS Defining Cancers (45) 29.8 * Chemotherapy, targeted therapy, drug-antibody conjugates, hormone therapy, radionuclide therapy.

  PublisherDOI

• Reliability of Large Language Model Knowledge Across Brand and Generic Cancer Drug Names

  JCO Clinical Cancer Informatics · 2025-06-01 · 1 citations

  articleOpen access

  PURPOSE To evaluate the performance and consistency of large language models (LLMs) across brand and generic oncology drug names in various clinical tasks, addressing concerns about potential fluctuations in LLM performance because of subtle phrasing differences that could affect patient care. METHODS This study evaluated three LLMs (GPT-3.5-turbo-0125, GPT-4-turbo, and GPT-4o) using drug names from HemOnc ontology. The assessment included 367 generic-to-brand and 2,516 brand-to-generic pairs, 1,000 drug-drug interaction (DDI) synthetic patient cases, and 2,438 immune-related adverse event (irAE) cases. LLMs were tested on drug name recognition, word association, DDI (DDI) detection, and irAE diagnosis using both brand and generic drug names. RESULTS LLMs demonstrated high accuracy in matching brand and generic names (GPT-4o: 97.38% for brand, 94.71% for generic, P < .01). However, they showed significant inconsistencies in word association tasks. GPT-3.5-turbo-0125 exhibited biases favoring brand names for effectiveness (odds ratio [OR], 1.43, P < .05) and being side-effect-free (OR, 1.76, P < .05). DDI detection accuracy was poor across all models (<26%), with no significant differences between brand and generic names. Sentiment analysis revealed significant differences, particularly in GPT-3.5-turbo-0125 (brand mean 0.67, generic mean 0.95, P < .01). Consistency in irAE diagnosis varied across models. CONCLUSION Despite high proficiency in name-matching, LLMs exhibit inconsistencies when processing brand versus generic drug names in more complex tasks. These findings highlight the need for increased awareness, improved robustness assessment methods, and the development of more consistent systems for handling nomenclature variations in clinical applications of LLMs.

  PublisherOA PDFDOI

• Supplemental Fig. 1 from The GENIE BPC NSCLC Cohort: A Real-World Repository Integrating Standardized Clinical and Genomic Data for 1,846 Patients with Non–Small Cell Lung Cancer

  2025-11-25

  articleOpen access

  <p>Supplemental Figure 1: OS and PFS for KRAS-mutated NSCLC treated with first-line platinum-based chemotherapy.</p>

  PublisherDOI

• Analysis of ASCO clinical guidelines authorship and institutional representation using HemOnc.org.

  Journal of Clinical Oncology · 2025-05-28

  articleSenior author

  e13577 Background: Evidence-based clinical practice guidelines (CPGs) are vital for safe and updated oncology care. ASCO produces CPGs across solid and blood cancers and supportive care. Guideline authors include volunteers and ASCO staff. This study characterizes ASCO CPGs by authorship and their affiliations. Methods: ASCO-only and collaborative CPGs published in peer-reviewed journals, excluding rapid recommendations, were analyzed. CPGs were categorized into 16 clinical groups. Author and institutional data were extracted from PubMed and normalized using Python and manual abstraction. Authors’ gender was determined using gender-guesser, genderAPI, and manual review. Statistical comparisons used Fisher’s exact test. Data were sourced from HemOnc.org on 12/02/2024. Results: We identified 146 eligible CPGs (1999–2024), with 1297 unique authors. 102 (8%) were ever-first authors and 108 (8%) were ever-last (senior) authors. The most prolific first author was in this role six times (breast/classical hematology), while the most prolific last author, five times (breast cancer). A total of 482 institutions were represented, and senior authorship spanned 59 institutions, with 33% affiliated with five institutions: University of Michigan, MD Anderson, Dana-Farber, Johns Hopkins, and ASCO. Only four CPGs had first and senior authors from the same institution. 844 (65%) authors were USA-based. Canada, UK, Italy, Japan, and Netherlands were sequentially the next five highest contributors (142 authors). Top groups included breast (36 CPGs), gastrointestinal (21), and supportive care (20). Immunotoxicity CPGs had the highest collaboration (31 authors/guideline) compared to the lowest in radiotherapy toxicity (11.7 authors/guideline). Over the 25-year publication period, 18 authors transitioned from first to senior authorship, and 48 different authors transitioned from middle to first or senior authorship. Based on algorithmically assigned gender, 57% of all authors, 57% of first authors, and 60% of senior authors were men. Immunotoxicity had the highest representation of women (60%), while radiotherapy toxicity had the lowest (10.5%). Of the 1297 authors, 283 (22%) also contributed to pivotal clinical trials publications supporting regulatory approvals. These authors were more likely to be men (70% men vs 30% women) than the non-contributing authors (53% men vs 47% women) (OR 2.02, 95% CI 1.52–2.68). Conclusions: ASCO CPGs demonstrate broad institutional and international representation, though senior author concentration in a smaller number of institutions was observed. We show gender differences exist in ASCO guideline authorship, although they are significantly less prominent overall than in the subgroup of pivotal clinical trial authors. Future research should explore whether ASCO guideline authorship diversity reflects the demographics of oncology subfields.

  PublisherDOI

• A deep learning model to predict glioma recurrence using integrated genomic and clinical data

  Communications Medicine · 2025-08-19 · 3 citations

  articleOpen access

  Gliomas account for approximately 25.5% of all primary brain and central nervous system (CNS) tumors and 80.8% of malignant brain and CNS tumors. The prognosis varies considerably; patients with low-grade gliomas (LGGs) have 5-year survival rates of up to 80%, while patients with higher-grade gliomas (HGGs) often experience rates below 5%. Recurrence is a common challenge, occurring in 52% to 62% of patients with LGGs and 90% of patients with HGGs, complicating clinical management and treatment planning. Currently, no widely available models exist for reliably predicting early glioma recurrence, which is critical for optimizing patient outcomes. Machine learning (ML) and deep learning (DL) techniques have shown promise in predicting recurrence for various cancers, with those utilizing multimodal data sources showing increasing promise. We developed a DL-based predictive model with attention mechanisms, gLioma recUrreNce Attention-based classifieR (LUNAR), to predict early vs. late glioma recurrence using clinical, mutation, and mRNA-expression data from patients with primary grade II-IV gliomas from The Cancer Genome Atlas (TCGA) and, as an external validation set, the Glioma Longitudinal Analysis Consortium (GLASS). Our model outperforms traditional ML models and non-attention counterparts, achieving area under the receiver operating characteristic curve (AUROC) of 82.84% and 82.54% on the TCGA and GLASS datasets, respectively. Our results demonstrate the potential of multimodal DL classifiers for predicting early glioma recurrence. By integrating clinical, mutational, and transcriptomic data from patients, LUNAR enables improved risk stratification. Its consistent performance across two independent datasets underscores its robustness. Gliomas are a type of brain tumor that often return after treatment, making them difficult to manage. Patients with low-grade gliomas tend to have better long-term survival, while high-grade gliomas are much more aggressive. Deep learning is a type of machine learning where computers utilize layered networks, known as neural networks, to recognize patterns in large datasets. We developed a deep learning model to predict which patients with glioma will have early recurrence. This model uses clinical data from patients with glioma, as well as information about patients’ genetic mutations and gene activity. The model was tested on two independent patient datasets and outperformed traditional prediction methods indicating its potential to enhance care and treatment planning for patients with glioma. Patricoski-Chavez et al. use a deep learning-based model with attention mechanisms that integrates clinical, mutation, and mRNA expression data from patients to predict early versus late glioma recurrence. The model outperforms traditional machine learning approaches across two independent datasets.

  PublisherOA PDFDOI

• Systemic Anticancer Therapy Timelines Extraction From Electronic Medical Records Text: Algorithm Development and Validation

  JMIR Bioinformatics and Biotechnology · 2025-07-07

  articleOpen access

  Background: The systemic treatment of cancer typically requires the use of multiple anticancer agents in combination or sequentially. Clinical narrative texts often contain extensive descriptions of the temporal sequencing of systemic anticancer therapy (SACT), setting up an important task that may be amenable to automated extraction of SACT timelines. Objective: We aimed to explore automatic methods for extracting patient-level SACT timelines from clinical narratives in the electronic medical records (EMRs). Methods: We used two datasets from two institutions: (1) a colorectal cancer (CRC) dataset including the entire EMR of the 199 patients in the THYME (Temporal Histories of Your Medical Event) dataset and (2) the 2024 ChemoTimelines shared task dataset including 149 patients with ovarian cancer, breast cancer, and melanoma. We explored finetuning smaller language models trained to attend to events and time expressions, and few-shot prompting of large language models (LLMs). Evaluation used the 2024 ChemoTimelines shared task configuration-Subtask1 involving the construction of SACT timelines from manually annotated SACT event and time expression mentions provided as input in addition to the patient's notes and Subtask2 requiring extraction of SACT timelines directly from the patient's notes. Results: Our task-specific finetuned EntityBERT model achieved 93% F1-score, outperforming the best results in Subtask1 of the 2024 ChemoTimelines shared task (90%). It ranked second in Subtask2. LLM (LLaMA2, LLaMA3.1, and Mixtral) performance lagged the task-specific finetuned model performance for both the THYME and shared task datasets. On the shared task datasets, the best LLM performance was 77% macro F1-score, 16% points lower than the task-specific finetuned system (Subtask1). Conclusions: In this paper, we explored approaches for patient-level timeline extraction through the SACT timeline extraction task. Our results and analysis add to the knowledge of extracting treatment timelines from EMR clinical narratives using language modeling methods.

  PublisherDOI

• A Field Guide to Deploying AI Agents in Clinical Practice

  arXiv (Cornell University) · 2025-09-30 · 1 citations

  preprintOpen access

  Large language models (LLMs) integrated into agent-driven workflows hold immense promise for healthcare, yet a significant gap exists between their potential and practical implementation within clinical settings. To address this, we present a practitioner-oriented field manual for deploying generative agents that use electronic health record (EHR) data. This guide is informed by our experience deploying the "irAE-Agent", an automated system to detect immune-related adverse events from clinical notes at Mass General Brigham, and by structured interviews with 21 clinicians, engineers, and informatics leaders involved in the project. Our analysis reveals a critical misalignment in clinical AI development: less than 20% of our effort was dedicated to prompt engineering and model development, while over 80% was consumed by the sociotechnical work of implementation. We distill this effort into five "heavy lifts": data integration, model validation, ensuring economic value, managing system drift, and governance. By providing actionable solutions for each of these challenges, this field manual shifts the focus from algorithmic development to the essential infrastructure and implementation work required to bridge the "valley of death" and successfully translate generative AI from pilot projects into routine clinical care.

  PublisherOA PDFDOI

• SA71 Integrating Next Generation Sequencing, EHR, and Claims Data to Extend Follow-Up in a Real-World Advanced Lung Adenocarcinoma Biomarker-Treatment Landscape

  Value in Health · 2025-07-01 · 1 citations

  article
  PublisherDOI

• Supplemental Fig. 2 from The GENIE BPC NSCLC Cohort: A Real-World Repository Integrating Standardized Clinical and Genomic Data for 1,846 Patients with Non–Small Cell Lung Cancer

  2025-11-25

  articleOpen access

  <p>Supplemental Figure 2: OS and PFS for patients with NSCLC treated with first-line platinum-based chemotherapy.</p>

  PublisherDOI

• Bringing Trustworthy Artificial Intelligence to the Clinical Forefront at <i>JCO</i> : A Guide for Studies Testing Artificial Intelligence Models

  Journal of Clinical Oncology · 2025-10-02 · 1 citations

  editorial
  PublisherDOI

Recent grants

• Enhancing the HemOnc Knowledgebase of Chemotherapy Drugs and Regimens

  NIH · $2.7M · 2022–2027

• Natural Language Processing Platform for Cancer Surveillance

  NIH · $2.1M · 2019–2024

• Natural Language Processing Platform for Cancer Surveillance

  NIH · $1.3M · 2019–2021

• Cancer Deep Phenotype Extraction from Electronic Medical Records (renewal)

  NIH · $5.3M · 2020–2030

• Network Analysis for a Data-Driven Approach to Cancer Care

  NIH · $1.2M · 2018–2022

Frequent coauthors

• Gregory J. Riely

  Memorial Sloan Kettering Cancer Center

  162 shared

• Deborah Schrag

  151 shared

• Nikolaus Schultz

  140 shared

• Michele L. Lenoue-Newton

  Vanderbilt University Medical Center

  135 shared

• Sanjay Mishra

  Brown University

  134 shared

• Philippe L. Bédard

  133 shared

• Kenneth L. Kehl

  132 shared

• Celeste Yu

  Princess Margaret Cancer Centre

  130 shared

Education

• M.D.

  Boston University

  2005

• M.S.

  University of California at San Diego

  2001

• B.S.

  Massachusetts Institute of Technology

  1999