clinTALL: machine learning-driven multimodal subtype classification and treatment outcome prediction in pediatric...

Area of Science:

• Computational biology and bioinformatics
• Machine learning in oncology
• Pediatric hematologic malignancies

Background:

• Childhood T-lineage acute lymphoblastic leukemia (T-ALL) is aggressive with poor prognosis.
• Existing risk stratification for T-ALL is insufficient compared to B-cell precursor ALL.
• Multi-omics data integration offers potential but faces interpretation challenges.

Purpose of the Study:

• To develop a deep learning pipeline (clinTALL) for pediatric T-ALL.
• To enable accurate subtype classification and treatment outcome prediction.
• To facilitate clinical translation of complex multi-omics data.

Main Methods:

• Developed clinTALL, a deep learning multi-task pipeline integrating multimodal data (clinical, genomic, transcriptomic).
• Employed a neural network architecture for shared latent embedding and multi-task prediction.
• Utilized a competing risk model for event-specific outcome prediction on 1309 pediatric T-ALL samples.

Main Results:

• Transcriptomic-only model achieved 92.2% accuracy for subtype prediction and 65.9% C-index for event-free survival (EFS).
• Multimodal integration maintained high subtype accuracy (91.7%) and improved EFS C-index to 67.5%.
• Accurate prediction of induction failure (96.0% C-index) and second malignant neoplasm (62.1% C-index) was achieved.

Conclusions:

• The clinTALL framework provides automated, accurate subtype classification and outcome inference using multimodal data.
• This advances precision risk stratification for pediatric T-ALL.
• clinTALL is available as a user-friendly Docker application.