Alzheimer's disease (AD) is a neurodegenerative disease that causes dementia in aging people. Drug repurposing is an efficient approach to accelerate the process of identifying drug candidates for AD. Single-cell RNA-sequencing technique provides comprehensive transcriptomic profiling of gene expression across multiple cell types. In this study, we developed ADPurpose, a gene-induced drug repurposing approach with deep learning-based drug-target interaction to propose drug candidates for AD. We analyzed over 1000 differentially expressed genes across six neuron cell types between AD and healthy people and generated a large pool of drug candidates. By drug-target interaction prediction, we ranked hundreds of repurposed drug candidates and finally generated the top ten drugs for each cell type. Among the top ten drugs derived from excitatory (Ex) and inhibitory (In) neurons, fifteen drugs were related to AD and six drugs have effects on AD, including chlortetracycline, dirithromycin, asiaticoside, fludrocortisone, valinomycin and ciprofibrate. ADPurpose shows high efficiency for rapid drug repurposing using transcriptomic information. ADPurpose is also transferable to drug repurposing studies for other diseases.