As in other fields, artificial intelligence (AI) and machine learning (ML) are transforming oncology and hematology practice.
While the terms AI and ML are used interchangeably in medicine, they actually have different roles in the fields of computer science and technology: “Artificial intelligence encompasses a diverse set of techniques and methodologies dedicated to creating systems capable of mimicking human-like intelligence and decision-making processes,” Mohamed ElHadary and colleagues write. Blood Review“Machine learning, on the other hand, is a specialization within the AI domain that focuses on developing models, called function approximators. These models can autonomously make informed decisions and draw conclusions by identifying patterns and extracting meaningful insights from raw data.”
An industry-sponsored study leveraging AI and ML to improve understanding of chronic lymphocytic leukemia (CLL) was presented in a poster by Bessi Qorri, PhD, MSc, and colleagues from NetraMark at the 2024 ASCO Annual Meeting.
CLL is a lymphoproliferative disorder characterized by the proliferation of monoclonal B cells and is the most common adult leukemia in Western countries, accounting for 25% to 30% of leukemias in the United States. The exact etiology of CLL is unknown, but genetic factors play a role. CLL is rarely seen in children, and its incidence in adults increases with age.
Diagnosis requires blood tests showing increased lymphocytes, peripheral blood smears, and flow cytometry. In recent years, advances in AI and ML have produced various models and algorithms that support the diagnosis and classification of CLL. [https://www.sciencedirect.com/science/article/pii/S0268960X23000954]
Advanced Treatment Decision-Making
As the researchers note in their summary, they applied NetraAI's proprietary AI and ML systems to a CLL dataset and used it to improve the analysis and interpretation of the corresponding transcriptomics data (GSE39411).
According to the NetraMark website, NetraAI is “an ML system that provides an intuitive interface for scientists to interact with multimodal datasets to discover associations such as efficacy, toxicity, and placebo response in smaller datasets.” The system generates hypotheses in the form of interactive representations of patient populations showing heterogeneity and statistically significant drivers. An automated agent interacts with the representations, taking into account all variables, sample subpopulations, and reports what it finds in the dataset.
Using this approach, the authors identified significantly lower levels of FADS3, GSDME, LPL, IMMT, NMB, and AEBP1, and higher expression of COBLL1, P2RY1R, PDE8A, SYNE2, and FCRL3 as characteristic of low-grade CLL, suggesting that these markers may be useful for less aggressive disease management strategies.
In 104 CLL samples, the researchers detected two distinct subgroups within malignant CLL distinguished by unique genetic markers: one group, consisting of 31 malignant CLL samples, is characterized primarily by the expression of lipoprotein lipase, while the other group, consisting of 22/23 malignant samples, is characterized by ZBTB20 and SYNE2.
“Our findings highlight the heterogeneity of aggressive CLL and the potential of these markers to guide prognostic assessment and treatment decisions,” the authors wrote. “This study not only improves our understanding of CLL heterogeneity, but also lays the foundation for the development of more personalized and effective treatments. Identifying the causes of aggressive and delayed disease progression paves the way for targeted therapies that could significantly improve outcomes for CLL patients.”
