Haemorrhagic stroke (HS) is a devastating form of stroke with a high fatality rate. The lack of rapid lesion detection limits early diagnosis of HS. Several susceptibility genes of HS found in genome-wide association studies (GWAS) warrant transcriptional-level biomarkers for useful utility. 13 GWAS level loci with minor allele frequency ≥ 0.05 were selected out of 95 loci related to HS. After validation, the mRNA expression in peripheral leukocytes of 11 genes (PMF1, SLC25A44, CASZ1, NINJ2, WNK1, DYRK1A, LRCH1, LDLR, SMARCA4, AQP9, and LIPC) were measured in the HS case-control study (64 HS cases vs. 128 controls), and then verified in an ischemic stroke (IS) case-control study (67 IS cases vs. 61 controls). LIPC (P=0.002) and CASZ1 (P=0.040) were downregulated in HS patients, while SLC25A44 was upregulated (P=0.009). In the IS case-control study, the differential expression of LIPC (P=0.034) and SLC25A44 (P<0.001) was observed. Although CASZ1 expression was not different between IS cases and controls (P=0.419) with fold change (FC) of 1.205, the direction of expression was opposite to that in HS case-control study (FC=0.747). The ROCtrfgs of traditional risk factors and gene score which was estimated by combining LIPC, CASZ1, and SLC25A44 expression weights, improved the utility for HS identification by 14.1% compared with ROCtrf (P=0.001). Expression of LIPC, CASZ1, and SLC25A44 could serve as potential biomarkers for identifying HS. CASZ1 might be a cause-specific indicator for differentiating HS from IS. Transcriptional score of these three genes could improve performance of the traditional risk model for HS identification.