How AI and Genetic Testing Are Improving Lymphoma Classification and Relapse Insights

Read findings presented at the 2025 ASH conference on how artificial intelligence (AI) and new genetic research may improve how lymphoma is classified and understood. These advances could help doctors make more informed treatment decisions for people living with lymphoma. 

AI tools may improve lymphoma classification from standard biopsy slides

A study introduced “LymphoVision,” an AI model trained to read pathology slides, which are stained tissue samples viewed under a microscope. The goal was to improve how lymphoma is classified and divided into subtypes, since this information can guide treatment choices and help estimate risk. 

The model did well in several important tasks:

• It identified the cell-of-origin type in diffuse large B-cell lymphoma (DLBCL). Cell of origin describes the type of normal B cell the lymphoma most closely resembles.
• It graded follicular lymphoma, which means it measured how quickly lymphoma cells grew.
• It classified several lymphoma subtypes, including DLBCL, follicular lymphoma, mantle cell lymphoma (MCL), marginal zone lymphoma (MZL), classical Hodgkin lymphoma (cHL), and non-cancerous conditions.

This research is important because correctly identifying the lymphoma subtype is the first step in choosing the best treatment plan. Artificial intelligence will not replace expert pathologists, who are doctors trained to examine tissue under a microscope. However, it may help make classifications more consistent and faster, especially in centers that do not have lymphoma specialists. 

Read this abstract: Lymphovision: A lymphoma-specialized foundation model for histology-based lymphoma classification and subtyping 

Relapsed DLBCL often keeps its original genetic “cluster,” which may explain relapse patterns

DLBCL is not just one condition. Researchers have grouped it into genetic “clusters” called C1 through C5. These clusters behave differently and may respond differently to the standard first treatment. 

In one study, researchers looked at detailed genetic testing from 122 samples of DLBCL that had returned after the first treatment. They used a tool called DLBclass, which is a program that sorts lymphoma into genetic groups based on DNA changes.

One key finding was that many of the genetic features linked to relapse were already present at diagnosis. About 74% of patients had the same cluster at diagnosis and at relapse. This suggests that, in many cases, the biology of the lymphoma is set early on.

However, some tumors changed over time. A few shifted toward a cluster linked to TP53 gene mutations and genomic instability. A TP53 mutation is a change in a gene that normally helps control cell growth. Genomic instability means the cancer cells have many DNA errors, which can make the lymphoma harder to control.

The study also found that certain clusters were linked to where the lymphoma came back. For example, relapses outside the lymph nodes, including those in the central nervous system, were more likely to be in the C5 cluster. Relapses in the lymph nodes were more often in the C3 cluster. 

Researchers noticed that the loss of immune-related genes was more common at relapse. These genes help the immune system target cancer cells. When they are lost, lymphoma cells may be better able to avoid the immune system. 

For people living with DLBCL, this research supports the value of genetic testing at diagnosis and sometimes again at relapse. Understanding the lymphoma’s genetic cluster may help refine risk estimates and guide participation in clinical trials that match treatment to the biology of the lymphoma. 

Read this abstract: Relapsed diffuse large B-cell lymphoma is shaped by molecular subtypes and genomic alterations 

Two subtypes inside “C5” DLBCL may point to different biology

C5 DLBCL is known as a higher-risk genetic group after the standard first treatment of R-CHOP. This study looked deeper and found two subtypes within C5: C5A and C5B.

C5A was mainly shaped by copy number changes, meaning there were extra or missing chromosome pieces. C5B was mainly shaped by specific gene mutations, including MYD88, L265P, and CD79B. Both subtypes had similarly poor progression-free and overall survival after R-CHOP compared with other DLBCL genetic groups. 

Over time, this research could influence patients with C5 subtypes to choose certain targeted therapies or clinical trials that better match their needs. 

Read this abstract: Characterization of two genetically distinct subtypes of C5 diffuse large B-cell lymphoma (DLBCL) 

Key takeaways 

Tools like AI slide analysis and genetic testing are helping doctors better define lymphoma subtypes and relapse patterns. Over time, the more precise information may better support personalized treatment plans and guide clinical trial options for people with lymphoma. 

Get the latest lymphoma updates delivered to you! The HealthTree newsletter shares core education, research advances, and more directly to your inbox. 

SIGN UP TODAY