This paper introduces the Conditional Prototypical Optimal Transport (CPOT) algorithm for clue identification in Multiple Choice Question Answering (MCQA) tasks. Existing clue-based methods suffer from inefficiencies, often relying on pseudo-labels or external resources, which introduce noise and additional computational demands. By contrast, the proposed CPOT method formulates clue identification as a sentence-oriented prototyping task, then further identifies sentences closest to prototype centroids as clues. Additionally, by leveraging the question and options as contextual guides, CPOT extends traditional Optimal Transport (OT) theory with constraints for unique assignment and uniform distribution across prototypes. This approach ensures semantically similar features converge within their prototypes and also maintains diversity among identified clues, enhancing answer accuracy. Empirical studies on several competitive benchmarks consistently demonstrate the superiority of our proposed method over different traditional approaches, with a substantial average improvement of 1.1—3.5 absolute percentage points in answering accuracy.