Multi-Cohort Meta-Analysis of Rnaseq Data to Determine Sex Bias in Pediatric Acute Myeloid Leukemia

Leukemia is the most common childhood cancer, and remains a leading cause of childhood cancer mortality. The most common leukemia in children, B-cell acute lymphoblastic leukemia, has a 92% cure rate (5-year overall survival). In contrast, pediatric acute myeloid leukemia (pAML) has a 68% cure rate, which has not changed significantly despite 40 years of trials.

AML exhibits some of the most striking sex-specific disparities in childhood cancer, affecting disease risk, therapy response, and transplant outcomes. pAML cohorts are consistently about 45% female and 55% male, yielding a male excess risk of 22%. Some aggressive subtypes, such as NUP98::NSD1 fusions seen in many treatment failures, the male:female ratio rises to 2:1, and to 3:1 near puberty. Up to 50% of pAML patients show loss of X or Y, with loss of Y (LOY) prognostic for lifetime all-cause mortality in males. Even mutation-targeted therapies such as midostaurin show significant differences in efficacy by sex in Phase III trials.

We sought to illuminate these differences via meta-analysis of bulk RNA sequencing (RNAseq) data from 2,000 TARGET (Therapeutically Applicable Research to Generate Effective Treatments) initiative patients and matched controls to identify sex-specific transcriptional differences. Subsequently, we interrogated these findings by performing a meta-analysis and cross-projection of publicly available single-cell RNA sequencing (scRNAseq) data from a variety of pediatric hematopoietic malignancies from eight published GEO datasets and two Alex's Lemonade Stand Foundation (ALSF) datasets, comprising 225 myeloid or myeloid-like pediatric leukemia patients and 23 non-leukemic specimens. Using Non-negative Matrix Factorization (NMF), we projected all scRNAseq data into the same latent space as the Chan Zuckerberg Initiative (CZI) human, matched-tissue scRNAseq. These NMF projections were then utilized to inform a comprehensive meta-analysis of both bulk and scRNAseq data from myeloid-like diseases. By leveraging both bulk and single-cell transcriptomic data, we aimed to elucidate the role of sex in the molecular underpinnings of pediatric hematopoietic malignancies, potentially unveiling novel therapeutic targets and contributing to personalized medicine approaches in these diseases.

Combined with the development of new models for predisposition conditions such as RUNX1-FPDMM, our work begins to reveal the molecular underpinnings of sex bias in treatment response and immunogenicity for pAML patients, and creates a foundation to better understand how genetic lesions interact with age and sex to shape risk in this lethal malignancy.

Triche:NIH: Research Funding; Aqtual: Current equity holder in private company; QuantumSI: Current equity holder in private company.