Background: Both stroke and depression are multi-factorial diseases, with both genetic and environmental factors likely to participate in their pathogenesis. Post stroke depression (PSD) is a common complication after stroke leading to poor functional outcome, increased physical disability and mortality. Although several genes have been associated with PSD, the genetic basis of PSD remains poorly understood. Method: A 2-stage candidate gene study by targeted sequencing was conducted involving stroke patients with or without depression and health controls. In the discovery stage (121 PSD, 131 non-PSD and 639 HC), logistic regression was used to WA associations respectively in PSD and non-PSD groups. In the replication stage (200 PSD, 218 non-PSD and 983 HC), 54 selected SNPs were again genotyped in an independent cohort. Fixed-effects inverse variance-weighted meta-analysis was used in the combined samples. Results: The study identified 2 novel genes associated with PSD [HTR3D (rs55674402, p = 0.002512, odds ratio (OR) = 0.7431); NEUROG3 (rs144643855, p = 0.00325, OR = 0.6523)] and 3 risk SNPs in one risk gene associated with non-PSD [PIK3C2B (rs17406271, p = 0.0006801, OR = 1.446; rs2271419, p = 0.0005836, OR = 1.497; rs2271420, p = 0.001031, OR = 1.431)] in the Chinese sample. NEUROG3 shows highest expression level in hippocampus. Functional enrichment analysis shows that susceptibility genes for PSD are mostly enriched in chemical synaptic transmission and regulation of lipid synthetic process. Limitations: The sample size was not sufficient to reach a genome-wide p value level. To overcome this shortage, some unique strategies were applied during the selection of SNPs for replication. Secondly, the age, gender composition and depressive severity between two stages were not well-matched. Different sample sources should be blamed, and to minimizing the influence, gender was corrected as co-variant in logistic regression. Conclusion: This study identified that HTR3D and NEUROG3 were linked with the susceptibility of PSD and PIK3C2B with stroke in the Chinese Han population. Further replication of these findings in a larger and better matched sample is warranted. Functional analysis suggests that the pathogenesis of PSD may be implicated in 5-HT synaptic transmission, neural plasticity and lipid metabolism, and therapeutic interventions targeting these pathways may be effective approaches for PSD treatment.