mRNA, a critical biomarker for various diseases and a promising target for cancer therapy, is central to biological and medical research. However, the development of multiplexed approaches for in situ monitoring of mRNA in live cells are limited by their reliance on enzyme-based signal amplification, challenges with in situ signal diffusion, and the complexity of nucleic acid design. In this study, we introduce a nonenzymatic catalytic DNA assembly (NEDA) technique to address these limitations. NEDA facilitates the precise in situ imaging of intracellular mRNA by assembling three free hairpin DNA amplifiers into a low-mobility, three-dimensional DNA spherical structure. This approach also enables the simultaneous detection of four distinct targets via the combination of fluorescent signals, with a detection limit as low as 141.2 pM for target mRNA. To enhance the efficiency of nucleic acid design, we employed computer-aided design (CAD) to rapidly generate feasible sequences for highly multiplexed detection. By integrating various machine learning algorithms, we achieved impressive accuracy of nearly 96.66% in distinguishing multiple cell types and 87.80% in identifying the same cell type under different drug stimulation conditions. Notably, our platform can also identify drug stimuli with similar mechanisms of action, highlighting its potential in drug development. This multiplexed 3D assembly sensing strategy with CAD not only enhances the ability to image nucleic acid sequences in situ simultaneously but also provides a novel platform for efficient molecular diagnostics and personalized therapy.