2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Monday, December 14, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Imaging, Omics Technologies, Computational Biology, Emerging Technologies, Gene Editing, Profiling, Treatment Considerations, Biological Therapies, Checkpoint Inhibitor, Women’s Health, Artificial Intelligence (AI), Machine Learning, Biological Processes, Pathogenesis, Immune Mechanism, Metabolism, Molecular Biology, Diseases, Bleeding and Clotting, Platelet Disorders, Thromboembolism, Thrombotic Disorders, Immunology, Research, Fundamental Science, Translational Research, Study Population, Human, Adult, Animal Model
PAPER

4835 Structure-activity based machine learning model identifies modifiers of estrogen-induced venous thromboembolism

Murat Yaman, MS, PhD, Lillian Smith2, Maya Sudhan3, Kennedy Gallaher4, Jiaming Yao, MS5, Sriram Chandrasekaran, PhD6, Jordan Shavit, MD, PhD7
2 University of Michigan, Ann Arbor, Michigan, United States
3 University of Michigan, Ann Arbor, Michigan, United States
4 University of Michigan, Ann Arbor, Michigan, United States
5 University of Michigan, Ann Arbor, Michigan, United States
6 University of Michigan, Ann Arbor, Michigan, United States
Venous thromboembolism (VTE), including deep vein thrombosis and pulmonary embolism, is the third leading cause of cardiovascular mortality in the U.S., affecting approximately 900,000 individuals and causing up to 100,000 deaths annually. Elevated estrogen from hormone therapy or contraceptives is a well-established acquired risk factor, yet the molecular mechanisms underlying estrogen-induced thrombosis remain largely unknown. Current therapies, such as warfarin and direct oral anticoagulants (DOACs), carry substantial bleeding risks exceeding seven events per 100 patient-years and can be fatal. DOACs alone account for nearly 40% of emergency visits related to oral anticoagulants, underscoring the need for safer therapies. In prior zebrafish FDA-repurposing screens, we found that proton pump inhibitors (PPIs) were antithrombotic at higher concentrations, while lower clinically relevant doses were prothrombotic. We evaluated patients on oral contraceptives and PPIs in two patient electronic health record databases and the results correlated with our zebrafish findings. To identify the pathways mediating the antithrombotic effects, we performed in silico target prediction and structure-activity-based cheminformatics analyses using M2D2, an open-source hybrid machine learning and systems biology approach we previously developed to screen drug targets across the entire proteome. This platform uses computationally generated drug-protein interaction data, thus precluding the need for costly omic studies. M2D2 utilized the chemical structures and antithrombotic activity levels of PPIs and other drugs from our prior zebrafish FDA-repurposing study to pinpoint potential mechanistic targets of PPIs. These analyses narrowed a proteome-wide candidate space to cytochrome b-245 alpha (CYBA), histamine H3 receptor (HRH3), serotonin receptors (HTR2A, HTR2B), and vascular endothelial growth factor receptor (VEGFR) family members as potential mediators. To confirm the functional roles of these targets, we combined CRISPR-Cas9 gene editing with pharmacological modulation in transgenic zebrafish expressing GFP-tagged fibrinogen. Larvae were treated with the synthetic estrogen mestranol with or without antithrombotic doses of the PPI esomeprazole, and thrombosis phenotypes were assessed in the venous system at 5 days post-fertilization. All analyses were blinded and evaluated using ordinal logistic regression with correction for multiple testing. We found that knockdown of serotonin receptor htr2b, and to a lesser extent histamine receptor hrh3, reduced PPI-mediated protection against estrogen-induced thrombosis. This effect was specific to the presence of PPIs, and there was minimal influence on thrombosis in their absence. In contrast, knockdown of other serotonin receptors (htr2a), cyba, and combinations of vegfrs increased estrogen-induced thrombosis. CYBA (p22phox) is an essential structural subunit of NADPH oxidase (NOX) complexes. Consistent with the prothrombotic effect of its knockdown, phorbol 12-myristate 13-acetate, a known NOX activator, was antithrombotic. This implicates classical NOX complexes (NOX1, NOX2, NOX4) and excludes CYBA-independent enzymes such as NOX5 and Duox. In contrast, isoform-specific inhibitors (GKT137831 for NOX1/4; diapocynin for NOX2) reduced estrogen-induced thrombosis, suggesting compensatory signaling upon partial NOX blockade or potential off-target mechanisms. In summary, cheminformatic analyses identified biologically relevant candidates with potential roles in PPI modulation of estrogen-induced VTE, which we have functionally validated in an in vivo vertebrate model. Ongoing work will determine whether all of these interactions are specific to the role of PPIs in modifying VTE, or are the results of more general effects on thrombosis. Further studies will also clarify isoform-specific NOX signaling dynamics. Together, these studies advance understanding of the mechanisms underlying estrogen-induced VTE and support the development of safer therapeutic strategies.
Jordan Shavit, MD, PhD: Genentech - Consultancy (Includes expert testimony); Sanofi - Consultancy (Includes expert testimony); Biomarin - Consultancy (Includes expert testimony); Pfizer - Consultancy (Includes expert testimony); CSL Behring - Consultancy (Includes expert testimony); NovoNordisk - Consultancy (Includes expert testimony); Takeda - Consultancy (Includes expert testimony); Medexus - Consultancy (Includes expert testimony)