Aligning preclinical AML models with immunotherapy development: principles for model selection

Advances in immunotherapy for acute myeloid leukemia (AML) have revealed critical gaps in selection of appropriate preclinical models. Conventional cytotoxic and targeted therapies could be tested in relatively straightforward systems. Immunotherapies are inherently distinct from conventional therapies, as their activity is shaped by dynamic, context-dependent interactions between leukemic cells, immune effectors, and the bone marrow microenvironment that are difficult to replicate outside the patient. Failure to recapitulate these interactions may limit the clinical translatability of promising results. Recognizing these shortcomings, the field has moved through several generations of modeling platforms. Existing platforms capture distinct but incomplete aspects of AML biology: in vitro systems offer control but lack immune context, syngeneic models provide intact immunity but limited human relevance, patient-derived xenograft (PDX) models preserve patient biology but lack immunity, and humanized models partially integrate both but remain constrained. In this review, we trace the development of these systems and use that trajectory to build a practical framework for model selection in AML immunotherapy research. Translational relevance depends on selecting preclinical models that align with the specific therapeutic question, rather than relying on availability or perceived complexity. To operationalize this, we propose a mechanism-driven framework that maps therapeutic mechanisms to the most appropriate biological contexts across preclinical platforms. This framework progresses sequentially from controlled mechanistic studies to in vivo validation, guiding model selection at each stage. Instead of prioritizing any single system, it emphasizes a complementary, question-driven approach that leverages the distinct strengths of each model while accounting for their limitations.