Data-driven hallmarks of cancer as a new paradigm for precision medicine: multi-omics and functional profiling in acute myeloid leukemia

The cancer hallmark concept defines biological properties that play a key role in cancer development and progression. The hallmarks reflect generic properties across all cancer types, and are not directly quantifiable nor applicable in cancer diagnostics. Here, we applied a data-driven approach to identify quantifiable hallmarks to be used as a basis for precision cancer medicine (PCM) in Acute Myeloid Leukemia (AML). We applied deep exome, transcriptome, DNA methylation and proteome profiling as well as ex-vivo functional testing of 525 drugs to 118 AML patient samples. Unsupervised multi-omic dimensionality reduction defined 11 independent axes of biological variability that we interpreted to reflect data-driven hallmarks (DDHM) of AML. Each DDHM integrates and ranks different data types and features, pinpointing those molecular features that were most informative for each hallmark. We calculated values for the 11 DDHM for each patient, constructing a DDHM-based precision medicine approach for AML diagnosis and therapy assignment. Most DDHMs were driven by other data types than genomics. For several DDHMs, different cytogenetic and mutational drivers converge on the same hallmarks and specific drug vulnerabilities. We also see how DDHM predictors of poor prognosis and high-risk AML are distinct from those that dictate specific drug response vulnerabilities. DDHM 2 reflected the cell differentiation path leading towards resistance to BH3 mimetics, ACK-inhibitors, and anthracyclines, and sensitivity to MEK inhibitors and TLR8 agonists. DDHM 2 derived a strong impact from protein expression of GATA2, MLLT11, DNMT3B and from RNA expression of WT1 and SOX4 as well as from multiple markers of different monocytic subtypes. DDHM 5 depicted cell cycle regulation and the Megakaryocytic-erythrocyte progenitor cell state that was enriched in AML patients with antecedent hematological disease. Interestingly, DDHM5 was linked to the sensitivity to purine analogs and vinca alkaloids. Moreover, DDHMs 1 and 8 captured clinical risk groups, prognosis, and responsiveness to hypomethylating agents. These DDHMs also revealed the phenotypic and functional biology that distinguished patients with high versus low FLT3-ITD mutant allele frequency. Validation of the DDHMs is realized through profiling and analysis of prospective AML samples, AML cell lines and previously published datasets. In summary, we present a data-driven approach for defining hallmarks in AML. The application of the DDHMs in AML provides a new paradigm for PCM and an opportunity for combinatorial therapeutic targeting. Each patient is characterized by a combination of independent and potentially druggable hallmarks, as opposed to traditional stratification in PCM, where each patient is assigned to one specific subgroup defined by genetic or other biomarkers. Citation Format: Tom Erkers, Nona Struyf, Tojo James, Francesco Marabita, Mattias Vesterlund, Nghia Vu, Cornelia Arnroth, Albin Österroos, Anna Bohlin, Sofia Bengtzén, Matthias Stahl, Rozbeh Jafari, Lukas Orre, Yudi Pawitan, Brinton Seashore-Ludlow, Janne Lehtiö, Sören Lehmann, Päivi Östling, Olli Kallioniemi. Data-driven hallmarks of cancer as a new paradigm for precision medicine: multi-omics and functional profiling in acute myeloid leukemia. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6612.