2 Clarke Drive
Suite 100
Cranbury, NJ 08512
© 2024 MJH Life Sciences™ and OncLive - Clinical Oncology News, Cancer Expert Insights. All rights reserved.
James K. McCloskey, MD, discusses the potential for machine learning to replace flow cytometry analyses in patients with hematologic neoplasms.
James K. McCloskey, MD, interim chief, Division of Leukemia, John Theurer Cancer Center, Hackensack Meridian Health, discusses findings from a study on next-generation sequencing (NGS) of flow cytometry CD markers, as well as the potential for machine learning to replace flow cytometry analyses in patients with hematologic neoplasms.
Although flow cytometry is widely used to diagnose patients with hematologic malignancies, the analysis of the data generated by this approach is subjective, and several neoplasms are difficult to diagnose using flow cytometry alone. Machine learning algorithms may compensate for the loss of subclonal analysis that occurs during the joint sequencing of cells in NGS. McCloskey and colleagues conducted a study to explore the potential use of a machine learning algorithm alongside RNA levels of CD markers to differentiate diagnoses between patients with various types of hematologic malignancies.
This study included 369 RNA samples from normal people to serve as a control and approximately 400 samples from patients with a variety of hematologic neoplasms. The malignancies diagnosed included acute myeloid leukemia (AML; n = 172), myelodysplastic syndrome (MDS; n = 218), myeloproliferative neoplasms (n = 68), acute lymphoblastic leukemia (n = 93), chronic lymphocytic leukemia (n = 74), mantle cell lymphoma (n = 38), and multiple myeloma (n = 83).
This study demonstrated that RNA sequencing of CD markers with the assistance of a machine learning computer model could reliably diagnose patients with hematologic malignancies, McCloskey said. Furthermore, this diagnostic method could differentiate between patients with hematologic neoplasms and healthy people. It could also differentiate between historically challenging diagnoses including AML and MDS with a high degree of specificity and sensitivity, McCloskey explained. Although the methods for diagnosing patients with diseases such as AML and MDS are evolving, this analysis provides insights into the potential role of machine learning in this area, McCloskey noted. Importantly, some patients who are diagnosed with MDS per World Health Organization criteria might exhibit disease characteristics more similar to that of AML, McCloskey concluded.
Related Content: