Acute myeloid leukemia (AML) is characterized by profound biological and clinical heterogeneity, where traditional mutation-centric frameworks often fail to fully capture the complex interplay between genomic drivers and functional phenotypes. In this study, we employed Multi-Omic Factor Analysis (MOFA) to integrate transcriptomic, mutational, pharmacological, and clinical data from a focused cohort of treatment-naïve de novo AML specimens. By utilizing a Bayesian framework, we identified nine latent factors that collapse high-dimensional data into distinct biological axes, explaining the majority of variance in the transcriptomic (54.5%) and clinical (23.7%) modalities. Our results characterized Factor 1 as a monocytic differentiation axis defined by high expression of mature myeloid markers such as CD14 and S100A8/9. Furthermore, the integration of MOFA-derived factors with the European LeukemiaNet (ELN) 2022 risk classification improved predictive accuracy, increasing the Harrell’s C-index from 0.66 to 0.72. These findings conclude that "molecularly silent" variance—biology not captured by somatic mutations alone—is a critical determinant of chemotherapy response and clinical outcome. Ultimately, this work provides a robust framework for transitioning toward a functional, multi-omic approach for personalized therapeutic selection and more precise risk assessment in AML.
Bahar Tercan (Fri,) studied this question.