Drug combinations often deliver greater efficacy and lower toxicity than single agents, but their molecular mechanisms of action (MoA) are frequently incompletely characterized. This study aimed to elucidate the MoA of a statin (atorvastatin or simvastatin) combined with ezetimibe by integrating drug‑induced transcriptomic data with computational modeling and network biology.
Two complementary transcriptomic sources were analyzed. First, RNA‑seq data were obtained from human hepatocyte‑like SOAT2‑only‑HepG2 cells treated with the drugs. Second, liver biopsy transcriptomes came from non‑obese normolipidemic patients with uncomplicated cholesterol gallstone disease enrolled in the Stockholm Study. These datasets enabled interrogation of drug responses in both an experimental hepatic cell model and human liver tissue.
The authors developed a novel Boolean logical modeling framework to simulate combinatorial MoA. The framework used fourteen two‑variable Boolean models to represent idealized gene expression patterns corresponding to different combinatorial modes. Each Boolean template encoded a distinct two‑input logic behavior (for example, additive, synergistic, antagonistic, and other combinatorial relationships), producing idealized differential expression templates that serve as reference signatures for matching real drug‑induced transcriptional responses.
Drug‑induced differentially expressed genes (DEGs) were identified in both datasets. In SOAT2‑only‑HepG2 cells, 1,560 genes were differentially expressed in at least one treatment condition. In the liver biopsy cohort, 565 genes were differentially expressed in at least one treatment condition. The study applied a pattern‑matching approach to associate observed DEGs with the idealized differential expression templates derived from the Boolean models. This template matching assigned downstream genes to specific combinatorial modes, enabling systematic interpretation of how the statin and ezetimibe interact at the transcriptional level.
After assigning genes to combinatorial modes, the downstream gene sets for each mode were mapped onto the human protein‑protein interactome. This network mapping step was used to identify connected pathways and network neighborhoods that are potentially modulated by each combinatorial mode. The approach links transcriptomic perturbations to network context, helping to reveal pathways relevant to therapeutic effects and to distinguish direct from downstream consequences of drug combinations.
The downstream genes associated with each combinatorial mode were subjected to functional enrichment and disease‑association analyses. These analyses were used to characterize biological pathways and disease processes enriched among the mode‑specific downstream genes, providing insight into additional therapeutic actions of statin–ezetimibe combinations beyond cholesterol lowering. The abstract reports that such analyses yield critical insights but does not enumerate specific pathways or disease terms in the summary.
Using this integrated pipeline, the authors report detection of both expected and novel combinatorial modes between statins and ezetimibe across the two datasets. The combination‑specific downstream gene sets and their mapping to the interactome revealed underlying pathways important for understanding the therapeutic effects of the drug combinations. The study highlights that transcriptomes induced by drug treatment, when combined with Boolean logical modeling and network mapping, can be informative for deciphering MoA of drug combinations.
This work presents a computational strategy that combines drug‑treated transcriptome data, a set of fourteen two‑variable Boolean logical templates, template matching, and human interactome mapping to infer combinatorial MoA. Key quantitative outputs reported in the abstract include 1,560 DEGs from SOAT2‑only‑HepG2 cells and 565 DEGs from human liver biopsies. The authors conclude that this integrated approach identifies both anticipated and previously unrecognized combinatorial behaviors and that downstream network and enrichment analyses provide useful biological and disease‑relevant context.
The abstract summarizes methods and high‑level results but does not provide detailed lists of specific pathways, the full set of genes assigned to each combinatorial mode, statistical thresholds used for DEG calling, or quantitative measures of model fit or template‑matching performance. Those details were not reported in the abstract and would require consulting the full text for complete methodological and result transparency.
This study demonstrates a systematic, network‑aware computational pipeline to interpret drug combination transcriptomes. For combination therapies such as statin plus ezetimibe, integrating Boolean logical modeling with transcriptomic and interactome data can reveal mode‑specific downstream genes and implicated pathways, informing mechanistic understanding and potential additional therapeutic targets.