To verify the specific differentiated subsets of monocytes in sepsis, and to screen and construct the differential gene set of monocytes used for early diagnosis of sepsis. Patients with sepsis admitted to Guangdong Provincial People's Hospital from June 2020 to March 2021 were enrolled, and peripheral blood mononuclear cells (PBMC) were extracted. Single-cell sequencing technology and pseudo-time analysis were used to verify the differential subsets of monocytes. Bioinformatics methods were used to analyze the expression of genes in differential subsets of monocytes and screen out differential genes for the preliminary construction of a candidate differential gene set. The digital polymerase chain reaction (PCR) technology was used to verify the candidate differential genes in PBMC of sepsis patients and sepsis human myeloid leukemia mononuclear cells (THP-1) models, and the Venn diagram was used to construct the final differential gene set of monocytes. Gene Expression Omnibus (GEO) database was used to validate the differential gene set of monocytes. (1) The results of cell annotation and pseudo-time analysis showed that the differentiation of NEAT1+CD163+ monocyte occurred in the early stage of sepsis was significantly different from other subsets, which validated that NEAT1+CD163+ monocyte was the characteristic subset in the pathological process of sepsis. (2) Twenty-two differential genes related to sepsis were screened out from the gene expression of NEAT1+CD163+ monocyte. After further verification by digital PCR, basic leucine zipper ATF-like transcription factor (BATF), JUNB proto-oncogene, carcinoembryonic antigen-related cell adhesion molecule 4 (CEACAM4), chromosome 9 open reading frame 95 (C9orf95), G protein subunit alpha 15 (GNA15), complement C3a receptor 1 (C3AR1), transforming growth factor beta 1 (TGFB1) and mitochondrial carrier homolog 1 (MTCH1) were screened out to construct the final differential gene set of monocytes. (3) The external validation results showed that C9orf95 gene had no data in GSE154918 and GSE133822 from GEO, it was excluded during validation. In GSE154918, the expressions of BATF, JUNB, CEACAM4, GNA15, C3AR1, TGFB1, and MTCH1 in the sepsis group were significantly higher than those in the healthy control group (log2expression level: BATF was 12.78±0.08 vs. 11.39±0.35, JUNB was 16.88±0.07 vs. 16.04±0.03, CEACAM4 was 14.73±0.08 vs. 13.77±0.05, GNA15 was 13.16±0.06 vs. 12.30±0.04, C3AR1 was 14.62±0.13 vs. 12.87±0.05, TGFB1 was 16.95±0.05 vs. 16.57±0.36, MTCH1 was 14.80±0.02 vs. 14.61±0.15, all P < 0.05). In GSE133822, the expressions of BATF, CEACAM4, GNA15, and C3AR1 in the sepsis group were significantly higher than those in the health control group (log2expression level: BATF was 8.66±0.16 vs. 7.92±0.14, CEACAM4 was 9.20±0.16 vs. 8.36±0.20, GNA15 was 10.66±0.18 vs. 10.13±0.16, C3AR1 was 11.49±0.27 vs. 10.48±0.16, all P < 0.05), while the expressions of JUNB, TGFB1, and MTCH1 were not statistically different between two groups. The results of gene set variation analysis (GSVA) showed that the enrichment scores of monocytes differential gene set of sepsis group were significantly higher than those of the healthy control group in both GSE154918 (0.38±0.04 vs. -0.44±0.02) and GSE133822 (0.56±0.02 vs. 0.20±0.05, both P < 0.01). Receiver operator characteristic curve (ROC curve) analysis showed that the differential gene set of monocytes had a reliable diagnostic value for early sepsis with the area under ROC curve (AUC) of 0.993 [95% confidence interval (95%CI) was 0.980-1.000] in GSE154918 and 0.944 (95%CI was 0.873-1.000) in GSE133822. A differential gene set of monocytes (BATF, JUNB, CEACAM4, GNA15, C3AR1, TGFB1, and MTCH1) screened out by single-cell sequencing and digital PCR technology has a reliable diagnostic value for the early sepsis, and may provide a new idea for the early diagnosis of sepsis.