IntegrateALL: An end-to-end RNA-seq analysis pipeline for multilevel data extraction and interpretable subtype...

Area of Science:

• Genomics
• Bioinformatics
• Oncology

Background:

• RNA-sequencing (RNA-seq) is crucial for diagnosing B-cell precursor acute lymphoblastic leukemia (B-ALL).
• Existing expression-based classifiers achieve high accuracy (~95%) but lack reproducible end-to-end solutions integrating genomic drivers and virtual karyotyping.

Purpose of the Study:

• To develop IntegrateALL, a standardized Snakemake pipeline for comprehensive RNA-seq analysis in B-ALL.
• To integrate expression-based subtype prediction, gene fusion/SNV calling, and virtual karyotyping for robust molecular characterization.

Main Methods:

• Developed IntegrateALL, a Snakemake pipeline for end-to-end RNA-seq analysis from FASTQ to subtype assignment.
• Introduced KaryALL, a machine learning classifier for distinguishing B-ALL ploidy subtypes using expression and minor allele frequency (RNASeqCNV).
• Validated RNA-based karyotyping against SNP-array data.

Main Results:

• IntegrateALL achieved unambiguous subtype assignments in 81.5% of 774 B-ALL cases, integrating gene expression with defining drivers.
• KaryALL demonstrated high accuracy (0.98) and F1 score (0.96) for ploidy classification on 615 independent test samples.
• Analysis of 1210 patients revealed 2.6% harbored dual subtype-defining drivers, including unexpected combinations.

Conclusions:

• IntegrateALL provides an adaptable, reproducible workflow for molecular B-ALL characterization.
• The pipeline systematically integrates genomic drivers and gene regulation for improved diagnostic accuracy.
• Findings highlight the complexity of B-ALL pathogenesis, including dual-driver events and potential oncogenic hierarchies.