Synthetic lethality has become an important strategy for exploiting selective vulnerabilities created by cancer-associated molecular alterations. Yet the therapeutic expression of a synthetic-lethal relationship is often less stable than a simple pairwise genetic model implies. Dependency can vary with cellular background, compensatory paralog activity, metabolic configuration, treatment exposure, and the experimental conditions used to measure fitness. This theoretical article develops an evidence-bounded account of synthetic lethality as a conditional cellular state rather than solely a fixed property of two genes or pathways. The analysis distinguishes a candidate synthetic-lethal interaction from the realized vulnerability that emerges when compensatory capacity is insufficient in a particular biological and treatment context. It further separates genuine biological conditionality from variation caused by assay design or model choice. Paralog buffering provides a mechanistically tractable example of how nominally similar lesions can produce different dependencies, while metabolic compensation and treatment-induced plasticity provide temporally dynamic routes through which vulnerability may persist, weaken, or re-emerge. The resulting account generates testable expectations for dependency mapping, biomarker development, pharmacological validation, and resistance research without assuming that all synthetic-lethal interactions are equally plastic or context sensitive.
