Multiomic state-transition framework reveals chemotherapy-induced metabolic reprogramming in acute myeloid leukemia

Acute myeloid leukemia (AML) is a highly lethal hematological malignancy with a long-term survival below 32%, largely due to disease relapse driven by treatment-resistant leukemia stem cells and metabolic reprograming. This poor prognosis underscores the need for frameworks that can predict disease dynamics and treatment effects. We previously applied state-transition theory to model AML disease evolution as trajectories of the mRNA and microRNA (miRNA) transcriptomes in their respective AML state-space, characterized by a leukemogenic potential with three critical points representing health, transition, and leukemia states. Here, we apply state-transition theory to characterize chemotherapy-induced transcriptomic and metabolic changes and test the hypothesis that chemotherapy alters the leukemogenic potential landscape by promoting transitions towards health while inducing metabolic shifts. Using a CBFB::MYH11 knock-in murine model of AML, we collected weekly peripheral blood samples before and after a “5+3” chemotherapy regimen consisting of cytarabine and daunorubicin, modeling the standard-of-care “7+3” regimen. All blood samples were subjected to bulk RNA-seq and miRNA-seq and we used PCA to construct mRNA- and miRNA-based state-spaces. Analyzing the mRNA and miRNA transcriptome trajectories over time, we observed that both transcriptome trajectories transitioned towards a health state post-chemotherapy but ultimately relapsed. Notably, the miRNA trajectories exhibited a delayed response of 2 or more weeks after treatment, revealing desynchronized mRNA-miRNA dynamics. To assess metabolic implications, we performed a Gene Set Variation Analysis (GSVA). This analysis revealed that oxidative phosphorylation, glycolysis and fatty acid metabolism pathways were downregulated during remission, indicating a low metabolic state, but were upregulated during relapse, consistent with metabolic reprogramming as a hallmark of AML recurrence. Finally, we extended the state-transition framework to two-dimensions (2D) to characterize how chemotherapy affects the mRNA-miRNA interplay and multiomic potential landscape. The multiomic state-space revealed a strong mRNA-miRNA correlation during AML progression, which is altered after chemotherapy. Further, the 2D state-transition model enables us to capture the effects of chemotherapy on multiomic potential and simulate how the mRNA-miRNA interplay changes over time. Importantly, model simulations capture response followed by relapse similar to the mRNA and miRNA time-series data trajectories. Together, our findings demonstrate that a mathematically grounded, multiomic state-transition framework captures chemotherapy-induced transcriptomic and metabolic dynamics, offering a systematic approach to predict response and identify metabolic vulnerabilities in AML.