A holistic characterisation of patients, which is key for progressing personalized oncology, requires the comprehensive integration of data from different molecular layers as well as clinical data. Machine learning can facilitate such statistically sound data integration strategy, but this comes with challenges around interpretability and trustworthiness. We demonstrate at the example of acute myeloid leukemia (AML) how interpretable machine learning-based multi-omics analyses reveal five proteogenomic AML subtypes, each reflecting specific biological features spanning genomic boundaries. Two of these proteomic subtypes correlate with patient outcome, but none are exclusively associated with specific genomic aberrations. Remarkably, one subtype (Mito-AML), is characterized by high expression of mitochondrial proteins and confers poor outcome, with reduced remission rate and shorter overall survival upon treatment with intensive induction chemotherapy. Functional analyses reveal that Mito-AML is more responsive to treatment with the BCL2 inhibitor venetoclax.

Download the slides for this talk.Download ( PDF, 5678.12 MB)