INTRODUCTION: This study aimed to identify diagnostic and therapeutic biomarkers related to glucose metabolism in sepsis, as hyperglycemia and blood glucose fluctuations influence sepsis progression. METHODS: Datasets from public databases were analyzed using various methods, including differential expression analysis, PPI network screening, machine learning algorithms and Mendelian randomization. A nomogram model was developed, and biomarker functions were explored through enrichment analysis, immunoinfiltration analysis, transcription factors (TFs) and microRNA (miRNA) prediction, and drug prediction. Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) was performed to validate the expression of biomarkers in sepsis and control group. RESULTS: There were 3,899 differential expressed genes (DEGs) in sepsis, with 141 related to glucose metabolism. Eleven hub genes were identified from the PPI network, and six biomarkers were selected through machine learning and area under the curve (AUC) validation. Notably, (OR = 1.0730, 95% CI: 1.0330-1.1160) and (OR = 0.9211, 95% CI: 0.8569-0.9902) had causal relationships with sepsis. The diagnostic nomogram based on these biomarkers showed good efficacy. Enrichment analysis suggested inhibits sepsis development, while promotes it. Drug prediction indicated strong interactions between and gigantol, and with echinatin. qRT-PCR showed reduced expression of and in sepsis, aligning with bioinformatics predictions. CONCLUSION: In summary, and are causally associated with sepsis, showing diagnostic potential. may inhibit sepsis development, while may promote it. These findings provide valuable insights for sepsis diagnosis and therapeutic drug development.