Abstract P25: Identification of biological components and novel clinical subsets of NPM1-mutated AML using an integrated, multi-omics approach

Acute myeloid leukemia is the most common acute leukemia in adults with a 5-year overall survival rate of only 28%. Relapse occurs in almost half of adults within 3 years. Due to a high degree of heterogeneity in biological and genetic features in AML tumor cells, current prognostic markers (including genetic aberrations) do not fully explain the ranges of phenotype and outcomes observed in AML patients. Recently, we characterized 13 DNA methylation signatures (epitypes) and a STAT hypomethylation signature (SHS) to be predictive of outcome and stable at relapse. Here, we used the integrative Multi-omics Factor Analysis (MOFA) approach to combine genetic, cytogenetic, transcriptomic, and our novel epigenetic information in an extensive multi-omics analysis to infer latent factors capable of explaining independent signatures comprising NPM1-mutated AML biology. The study used the well-annotated CALGB/Alliance AML cohort encompassing 581 NPM1-mutated patients. The analysis showed that RNA-sequencing contributed the most to explaining the variance between patients followed by RNA splicing, highlighting the depth of information present in the transcriptome. Surprisingly, genetic mutations contributed to only approximately 3% of the overall variance, while DNA methylation captured more than 15% of the total variance. We were able to infer 15 latent factors, 5 of which were associated with overall survival, event-free survival, and attainment of complete remission. All factors were significant after adjusting for established prognostic features, such as FLT3-ITD, sex, and age. Factors 2, 7 and 11 were associated with unfavorable outcomes, and variably included increased HOX signatures, suppressed TP53 activation, a stem cell-like phenotype, the epigenetic SHS signature, triple NPM1/FLT3-ITD/DNMT3A mutations, and alternative splicing, including NUP98. Factors 4 and 13 had a favorable prognostic value. Factor 13 uncovered a unique expression pattern of X-linked cancer testis antigen gene family members dividing NPM1 patients independent of sex, age, or common mutations. The factor was inversely associated with the expression of CD34, MN1 and BAALC – prognostic genes related to stemness. Analysis of immune cell subsets indicated differences in proportions of CD8 effector T cell populations between patients with high vs. low factor 13. This feature pattern and clinical significance were validated in an independent cohort (Beat-AML). Lastly, we clustered the 15 generated factors and identified 5 subsets of NPM1-mutated patients – 2 clusters with a favorable prognostic value, 2 clusters with unfavorable prognostic value, and one intermediate cluster. Overall, we identify 15 main sources of variance within NPM1-mutated patients, including 5 clinically-relevant factors associated with unique biological features, which together generate unique independent biomarker signatures not observed when individually considering separate omics data types.

Citation Format: Salma B. Abdelbaky, Kyoko Yamaguchi, Yue-Zhong Wu, Kevin R. Coombes, Lianbo Yu, Christopher C. Oakes. Identification of biological components and novel clinical subsets of NPM1-mutated AML using an integrated, multi-omics approach [abstract]. In: Proceedings of the Blood Cancer Discovery Symposium; 2024 Mar 4-6; Boston, MA. Philadelphia (PA): AACR; Blood Cancer Discov 2024;5(2_Suppl):Abstract nr P25.