Related Experiment Video
Updated: Jun 24, 2026
10:18
From a 2DE-Gel Spot to Protein Function: Lesson Learned From HS1 in Chronic Lymphocytic Leukemia
Published on: October 19, 2014
13.7K
DLBclass: a probabilistic molecular classifier to guide clinical investigation and practice in diffuse large B-cell
Björn Chapuy1,2,3,4, Timothy Wood5, Chip Stewart5
1Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA.
Blood
|December 16, 2024
Summary
A new classifier, DLBclass, accurately identifies 5 subtypes of Diffuse Large B-cell Lymphoma (DLBCL). This tool aids in classifying DLBCL cases for targeted therapies and clinical trials.
Area of Science:
- Oncology
- Genetics
- Bioinformatics
Background:
- Diffuse Large B-cell Lymphoma (DLBCL) is a heterogeneous cancer.
- Identifying molecular subtypes is crucial for targeted therapy.
- Previous work defined 5 DLBCL subtypes (C1-C5) based on genetic alterations.
Purpose of the Study:
- To validate the 5 DLBCL subtypes in an independent dataset.
- To develop a probabilistic molecular classifier for DLBCL subtypes.
- To confirm the classifier's performance on an independent test set.
Main Methods:
- Systematic comparison of machine learning models and feature reduction strategies.
- Development of a novel performance metric balancing accuracy and confidence.
- Training and validation of a neural network model (DLBclass) on 699 DLBCL cases.
Main Results:
- DLBclass achieved 91% accuracy on the training/validation set and 89% on the test set.
- For cases with confidence >0.7, DLBclass accuracy reached 97% (training/validation) and 98% (test).
- The classifier demonstrated robust performance in prospectively classifying DLBCL cases.
Conclusions:
- DLBclass provides a reliable method for prospective DLBCL classification.
- This tool facilitates the inclusion of DLBCL patients in genetically guided clinical trials.
- The framework supports the development of similar genomics-based classifiers for other cancers.
