DLBclass – A Breakthrough in Diagnosing and Managing Diffuse Large B-cell Lymphoma

Project Overview

• Despite being classified as a single disease under microscopic examination, diffuse large B-cell lymphoma (DLBCL) encompasses a variety of subtypes with distinct genetic profiles.
• A robust molecular classifier is critical to improving prognosis and tailoring therapeutic strategies effectively.
• Dana-Farber scientists and colleagues at the Broad Institute have developed DLBclass, a diagnostic tool designed to enhance the molecular diagnosis, prognostic accuracy, and treatment strategy for DLBCL.
• Dana-Farber seeks collaboration or licensing partners to commercialize the technology. 

Technology

Diffuse Large B-cell Lymphoma (DLBCL), the most common adult lymphoid malignancy, is known for its molecular heterogeneity, despite being a single diagnosis under the microscope. While over 60% of patients respond well to standard R-CHOP immunochemotherapy, the remainder either fail to respond or experience recurrence. The molecular heterogeneity of the cancer complicates treatment, necessitating a robust molecular classifier to better inform prognosis and therapeutic decisions. Toward that goal, a project led by Margaret Shipp, MD, Chair, Division of Hematologic Neoplasia at Dana-Farber and Professor of Medicine, Harvard Medical School, and Gad Getz, PhD, director of the Cancer Genome Computational Analysis Group at the Broad Institute have developed DLBclass, a new diagnostic tool aimed at improving DLBCL’s molecular diagnosis, prognostic accuracy, and treatment strategy for the disease. This innovation is poised to improve the molecular diagnosis for the 30,000 patients diagnosed with DLBCL annually and guide optimal therapy selection for improved response rates.  

In earlier research, Shipp and her team collaborated with colleagues from the Getz lab to define five distinct genetic subtypes of DLBCL, each with unique biological characteristics and varying responses to standard chemotherapy. This framework, published in Nature Medicine in 2018, laid the foundation for the development of a molecular classifier known as DLBclass by categorizing DLBCL cases into five distinct genetic subtypes that provide physicians with crucial insights into overall prognosis and treatment strategies. 

DLBclass is a sophisticated neural network-based probabilistic classifier that provides actionable genetic information in almost all DLBCL patients.

As described in the team’s recent paper in Blood, the team further validated taxonomy and developed a probabilistic molecular classifier, using a dataset of 699 primary DLBCL samples. The classifier’s algorithms analyze complex genetic data, identifying patterns and correlations that are indicative of specific subtypes. The researchers employed machine learning models to achieve high accuracy in classifying these subtypes, demonstrating 91% accuracy in training/validation sets and 89% in independent test sets. The classifier has potential for widespread clinical application, positioning it as a best-in-class diagnostic tool for DLBCL. 

The benefits of DLBclass are substantial. It enables precise classification of DLBCL subtypes, critical for enrolling patients in genetically guided clinical trials and optimizing treatment strategies. This classifier outperforms existing tools like the NIH’s LymphGen, which only classified 58% of cases into a significant portion of patients without clear subtype identification. In contrast, DLBclass successfully classified all samples, providing actionable genetic information for a larger number of patients. 

The goal of the project is to develop the test under CLIA guidelines, ensuring that the test produces consistent and accurate results across different conditions and equipment. Achieving CLIA certification would allow the test to be used in translational research and prospective clinical trials and practice, enabling oncologists to make informed treatment decisions based on a patient’s specific genetic subtype.

Team Members: Margaret A. Shipp, MD, Gad Getz, PhD, Björn Chapuy, MD, PhD, Chip Stewart, PhD

Applications

DLBclass has a wide range of applications in the clinical management of DLBCL. Its primary use is in the precise classification of DLBCL subtypes, which is crucial for enrolling patients in genetically guided clinical trials and determining optimal therapeutic choice by physicians caring for patients with DLBCL. By identifying the specific genetic subtype of a patient’s lymphoma, oncologists can make more informed decisions regarding treatment strategies, paving the way for improving patient outcomes. 

Furthermore, DLBclass serves as a framework for the development of genomic-based classification methods in other cancers, highlighting its potential beyond DLBCL.

Opportunities

Dana-Farber seeks collaboration or licensing partners to commercialize the technology and make it widely available for clinicians and members of the biotechnology community.  

Team Members

Margaret A. Shipp, MD

Chief, Division of Hematologic Neoplasia, Dana-Farber

Professor of Medicine, Harvard Medical School

Margaret's Bio

Gad Getz, PhD

Director, Cancer Genome Computational Analysis at the Broad Institute of MIT and Harvard

Director of bioinformatics in the Krantz Family Center for Cancer Research and Department of Pathology at Massachusetts General Hospital.

Gad's Bio

Björn Chapuy, MD, PhD

Former Research Scientist, Shipp Lab, Dana-Farber

Björn's Bio

Chip Stewart, PhD

Associate Director, Gad Getz lab, Broad Institute of MIT and Harvard

Chip's Bio