Integrating transcriptomic profiling and machine learning: A clinically actionable prognostic model for infant acute...

Area of Science:

• Pediatric Oncology
• Molecular Biology
• Bioinformatics

Background:

• Infant acute myeloid leukemia (AML) under 3 years old has poor prognosis and unique biology.
• Current risk stratification for infant AML requires age-specific assessment.

Purpose of the Study:

• To develop a simple RNA expression-based prognostic model for infant AML risk stratification.
• To refine treatment decisions for infant AML patients.

Main Methods:

• Utilized data from 213 infant AML patients from Children's Oncology Group trials.
• Applied machine-learning algorithms to develop the infant-prognostic-score (IPSscore/IPSgroup).
• Validated the model internally (127 cases) and externally (63 cases).

Main Results:

• The IPSgroup independently predicted outcomes in infant AML, outperforming existing RNA expression models.
• Identified high-risk infant AML patients who may benefit from hematopoietic stem cell transplantation (HSCT).
• Re-stratified 43% of infant AML patients into more accurate risk groups when combined with current systems.

Conclusions:

• The IPSgroup offers a valuable tool for clinical decision-making in infant AML.
• Incorporating gene expression analysis improves risk stratification and treatment strategies for infant AML.
• Infant AML exhibits significant clinical, molecular, and prognostic heterogeneity.