Novel Microfluidic and AI-Based Approach for Personalized Drug Selection in Chronic Myeloid Leukemia (CML) Patients

The management of chronic myeloid leukemia (CML) involves several FDA approved BCR::ABL1 tyrosine kinase inhibitors (TKIs), each with distinct efficacy and adverse events influenced by patient-specific biologic and genetic factors. Current genomic sequencing technologies aid in identifying optimal therapeutic options by correlating patient genotypes with clinical outcomes, increasingly supported by artificial intelligence (AI) in elucidating these correlations. Therefore, comprehensive assessments including in vitro assay remain crucial for the precise clinical prescription of anti-cancer therapies.

In this study, we introduce a pioneering microfluidic platform integrated with AI based image analysis to assist clinicians in selecting the most suitable BCR::ABL1 TKI for individual CML patients. Our system features a microfluidic cell culture array with 900 independent chambers, capable of generating nine distinct concentration gradients of six BCR::ABL1 TKIs at the same time. The array, combined with high-throughput AI-assisted image analysis, allows for the calculation of critical physiological metrics such as half maximal inhibitory concentration (IC50), growth rate inhibition (GR50), and drug sensitivity scores (DSS). These metrics provide a quantitative basis for comparing cellular drug sensitivities on a microfluidic chip, offering a robust scientific framework for personalized treatment strategies.

We conducted a comparative analysis of six BCR::ABL1 TKIs, including imatinib, nilotinib, bosutinib, ponatinib, dasatinib, and asciminib at predetermined concentration ranges, using the K562 CML cell line. The platform generated nine discrete concentration gradients within these ranges in a controllable and linear fashion, allowing characterization of cell viability across various TKI concentrations on a chip. This led to the development of comprehensive mathematical indices for assessing drug efficacy. Notably, the microfluidic platform demonstrated superior accuracy in determining IC50 values compared to traditional microplate assays. Specifically, the IC50 values of most TKIs quantified by the microfluidic platform closely matched reference values, with the exception of bosutinib, where IC50 values showed some variation depending on references. This accuracy is attributed to the platform's continuous, chemostat-like environment, which contrasts with the batch conditions of conventional methods, where cell growth is limited by initial conditions at the time of seeding.

Our findings revealed that the efficacy of the anticancer agents, ranked by DSS, is as follows: ponatinib, asciminib, dasatinib, imatinib, bosutinib, and nilotinib. Notably, ponatinib was the most effective on the K562 cell line, while nilotinib exhibited the least efficacy with significant variability depending on the concentration. It is emphasized that the same cellular image data in response to the TKIs can be quantified using other mathematical metrics with the aid of AI-assisted image processing algorithms. We are currently extending this approach to other cell lines and plan to present further experimental data at the conference. Furthermore, the platform allows us to investigate the effects of sequential drug dosing on cancer cell viability in a programmable manner, emulating time-dependent drug treatment schedules.

Future work involves applying this personalized TKI selection method to a clinical practice through prospective study, exploring the correlation between in vitro drug sensitivity and patient outcomes. This approach could significantly enhance the precision of treatment regimens, facilitating the transition towards personalized medicine in CML management. We will present our initial clinical trial results at the upcoming conference.

Kim:Pharmaessencia: Research Funding; Il-Yang: Honoraria, Research Funding, Speakers Bureau; Korea Otsuka: Honoraria, Research Funding, Speakers Bureau; Enliven: Honoraria, Research Funding; BMS: Honoraria, Research Funding, Speakers Bureau; Novartis: Honoraria, Research Funding, Speakers Bureau.