Artificial intelligence–empowered, clinically-integrated multiomics research in thrombosis: a call to action

1. Introduction

The past decade has marked the emergence of the multiomics era, driven by significant advances in high-throughput technologies across genomics, proteomics, metabolomics, and other omics. Innovations in hardware and software engineering have improved sensitivity and specificity, turnaround time, reduced costs, and increased accessibility, allowing for expanded global adoption of these platforms. Furthermore, major advances in computational biology, informatics, and advanced biostatistical techniques and modeling, including machine learning, have facilitated analytic capabilities across large multiomics datasets.

In parallel, recent years have witnessed the transformative power of artificial intelligence (AI). In translational research, AI has empowered investigators with the ability to combine comprehensive clinical patient and intervention data (eg, clinical test results, pharmacologic interventions, surgical procedures, and outcomes) with large-scale multiomics datasets. These integrative workflows have already led to substantial advances in disease prediction, stratification, and management.

The integration of multiomics data with AI allows researchers to identify patterns, predict outcomes, and generate important disease- and patient-specific insights that might not be realized through traditional analytical methods. Applications include the prediction of type 2 diabetes [1], assessment of pancreatic cancer risk in individuals with new-onset diabetes [2], identification of molecular subtypes and prognosis prediction in lung adenocarcinoma [3], and risk prediction of cardiovascular disease (CVD) [4].

These advances, combined with the realization that a single omics modality cannot solve many of the major clinical questions, have driven the rapid adoption of multiomics approaches, expanding the breadth and impact of the associated findings. By integrating genomics, proteomics, metabolomics, and other data types into a study, multiomics provides a more comprehensive assessment of putative biomarkers of disease and disease response, as well as a greater understanding of underlying pathophysiological processes. This knowledge, in turn, is critical for identifying key pathways in disease pathogenesis, as well as novel diagnostic markers and therapeutic targets.

To date, a large number of multiomics studies have been performed in the setting of cancer and CVD. These studies have advanced our understanding of the molecular signatures of type 2 diabetes and associated cardiovascular complications [5], oncogene-induced signaling [6], the identification of candidate biomarkers for patient stratification and therapeutic intervention in nonsmoking lung cancer patients [7], and disease progression in prostate cancer [8]. In the setting of CVD, multiomics studies have identified sex-specific signatures of major adverse cardiovascular events in individuals with type 2 diabetes [9], pathways connecting type 2 diabetes and atherosclerosis [10], and distinct multiomics signatures of multimorbidity of cardiometabolic diseases and cancer [11]. These examples, demonstrate the power of integrated multiomics approaches in enhancing patient care through improved diagnosis, precision treatment, and outcome prediction across multiple major diseases.

Despite the widespread adoption of multiomics approaches across biomedical disciplines, a paucity of research to date in the discipline of hematology, and specifically the field of thrombosis, has utilized an integrated multiomics approach. Specific examples from those few studies to date include investigations of: mechanisms of proinflammatory and prothrombotic functions in neutrophil-mediated inflammation and neutrophil extracellular trap formation in deep venous thrombosis (DVT) [12]; DVT risk stratification in trauma patients [13]; and identification of etiologic signatures in stroke [14].

While applications of AI to the integration of large clinical and omics datasets, as well as the integration of omics-driven findings into clinical decision-making tools (what we term “AI-empowered, clinically-integrated multiomics research”), is quickly becoming mainstream across numerous disciplines and disease areas, classical (ie, “nonmalignant”) hematology, and specifically thrombosis, have lagged behind as well. To date, only 3 published studies in the field of thrombosis have used AI in integrated multiomics work, focusing on the regulatory mechanisms of DVT [15], risk assessment of postoperative venous thromboembolism in cervical cancer patients [16], and the mechanism and diagnostic biomarkers for posttraumatic DVT in hypertensive patients after fracture [17]

Another piece of this puzzle, often missing in current multiomics investigations, is the integration of multiomics data with clinical data and nonomics laboratory assay data. For example, in thrombosis research, this could include data from functional assays and inflammatory markers. In the field of thrombosis, as knowledge of the interplay among coagulation, inflammation, and immunity continues to evolve, the value of combining these datasets in a single study cannot be overlooked.

2. Call to Action

The omics-AI revolution is here to stay. Multiomics and AI are essential tools that can revolutionize the way in which research findings are integrated, interpreted, and translated for improved patient outcomes. The thrombosis research community is perfectly positioned to take advantage of this transformation toward major improvements in patient outcomes. Achieving this will require deliberate cross-disciplinary collaboration and engagement with early-career researchers, who are best positioned to learn and adapt skills to the challenges that big data, machine learning, and AI. Access to high-quality, well-annotated datasets and appropriate analytical training is crucial. While some in our community are well-versed in the language of multiomics and AI, there are many for whom this represents and/or is a formidable learning curve. Broader engagement and education are essential to ensure equitable participation and sustained progress. Today, the opportunity for AI-empowered integration of multiomics data with well-curated, prospectively collected clinical research data, and the facilitation of the integration of omics-driven findings into clinical decision-making tools, lies before us, and with it, the opportunity to accelerate discovery and advancements in patient care in the field of thrombosis.

Today, the opportunity for AI-empowered, clinically-integrated, multiomics research lies before us, and with it, the opportunity to accelerate discovery and advancements in patient care in the field of thrombosis and hemostasis.