Pediatric acute myeloid leukemia (AML) is a severe hematological malignancy where allogeneic hematopoietic stem cell transplantation (allo-HSCT) serves as a critical, life-saving intervention. However, selecting the appropriate candidates for this intensive procedure remains a clinical challenge. Current clinical decision-making often relies heavily on minimal residual disease (MRD) testing, which can inadvertently introduce platform-specific biases and subjective clinical assessments.

To address this urgent need for objective evaluation, a new study published in Genes & Diseases by researchers from Chongqing Medical University, Sun Yat-Sen University, and Foshan University investigated a highly advanced transcriptomic approach. The researchers successfully developed HSCT-64, a novel parallel-risk framework designed to optimize precision transplantation for pediatric patients.

Workflow for the development, validation, and application of the HSCT-64 risk framework

A combination of four publicly available pediatric acute myeloid leukemia (pAML) cohorts was used as the discovery set to develop the HSCT-64 framework, which includes two parallel models: aHSCT-64 for allo-HSCT cases and nHSCT-64 for non-HSCT cases. Two independent pAML cohorts were employed for external validation. By comparing risk rankings generated by aHSCT-64 and nHSCT-64 in the independent cohort (n = 233), we identified the HSCT-benefiting subgroup (blue dots) as patients with a decreased risk rank from nHSCT-64 to aHSCT-64 and the HSCT-nonbenefiting subgroup (red dots) as those with an increased risk rank.

By exclusively utilizing comprehensive RNA-sequencing (RNA-seq) data, the research team constructed a powerful machine-learning model to evaluate individual patient transcriptomes. The robustness of this framework was rigorously tested across clinical datasets, featuring a large discovery cohort of 1,647 pediatric AML cases alongside a dedicated validation cohort of 223 patients from an independent Chinese cohort. The extensive bioinformatic data demonstrated that the HSCT-64 framework successfully and accurately identifies which pediatric patients will genuinely benefit from HSCT directly at the time of initial diagnosis.

Mechanistically, because HSCT-64 relies solely on RNA-seq-based gene expression profiles for prognosis, it overcomes the inherent biases introduced by traditional MRD testing platforms. This sophisticated approach minimizes human subjectivity in clinical assessments, providing a highly standardized and objective metric for evaluating disease severity and transplant suitability. By precisely stratifying patient risk and potential HSCT benefit, the model ensures that vulnerable patients receive critical stem cell transplants promptly, improving overall survival probabilities while shielding others from unnecessary transplant-related toxicities.

While these extensive data robustly highlight the critical advantage of utilizing a transcriptomic machine-learning framework to boost prognostic accuracy, continuous clinical integrations will further refine its global application.

In conclusion, implementing the HSCT-64 framework offers an advanced new strategy to refine precision clinical decision-making in pediatric oncology. This significant finding directly positions RNA-seq-based parallel-risk frameworks as highly compelling diagnostic tools, uniquely primed to deliver personalized and highly effective hematopoietic stem cell transplantation strategies for children battling acute myeloid leukemia.

Source – AlphaGalileo