Related Experiment Video
Updated: Sep 23, 2025
10:23
Single-cell Gene Expression Using Multiplex RT-qPCR to Characterize Heterogeneity of Rare Lymphoid Populations
Published on: January 19, 2017
11.1K
Multi-Cohort Transcriptomic Subtyping of B-Cell Acute Lymphoblastic Leukemia
Ville-Petteri Mäkinen1,2,3,4, Jacqueline Rehn5,6, James Breen6,7,8
1Computational and Systems Biology Program, Precision Medicine Theme, South Australian Health and Medical Research Institute, Adelaide, SA 5000, Australia.
International Journal of Molecular Sciences
|May 14, 2022
Summary
RNA sequencing accurately subtypes acute lymphoblastic leukemia (ALL) using machine learning. The Allspice R package aids clinical diagnosis by predicting ALL subtypes from mRNA-seq data.
Area of Science:
- Genomics
- Bioinformatics
- Oncology
Background:
- Acute lymphoblastic leukemia (ALL) subtyping is crucial for treatment.
- RNA sequencing (mRNA-seq) offers insights into genomic lesions driving ALL.
- Existing subtyping methods may lack clinical applicability or ease of interpretation.
Purpose of the Study:
- To develop machine learning models for ALL subtyping using mRNA-seq data.
- To identify robust transcriptome-wide biomarkers for ALL classification.
- To create an accessible tool for clinical settings to predict ALL subtypes.
Main Methods:
- Utilized a large training dataset (1279 ALL patients) and a validation cohort (767 ALL patients).
- Applied machine learning to mRNA-seq profiles to identify diagnostic associations.
- Developed and applied a novel batch correction method to account for cohort variations.
- Introduced the Allspice R package for predicting ALL subtypes from raw mRNA-seq counts.
Main Results:
- Robustly detected six ALL subtypes (ETV6::RUNX1, KMT2A, DUX4, PAX5 P80R, TCF3::PBX1, ZNF384) with high accuracy (PPV ≥ 87%).
- Distinguished five additional subtypes (CRLF2, hypodiploid, hyperdiploid, PAX5 alterations, Ph-positive) with moderate accuracy (PPV 52%-73%).
- Identified and excluded 64% of genes confounded by cohort effects using a novel batch correction method.
Conclusions:
- mRNA-seq profiles, when analyzed with appropriate methods, can reliably subtype ALL.
- The Allspice R package provides an easy-to-interpret tool for clinical ALL subtyping.
- This approach enhances diagnostic capabilities for ALL in real-world clinical settings.
