Objective Differentially expressed genes in oral squamous cell carcinoma (OSCC) were subjected to bioinformatic and Mendelian randomization analyses to elucidate their prognostic significance in OSCC. Methods The TCGA database and dataset GSE138206 were used to screen the common differential genes of OSCC, and their relationship was analyzed by using Mendelian randomization. The prognostic value of differential genes was further analyzed by Cox risk regression. The biological function of genes with high prognostic value was further evaluated by single gene differential analysis. Results A total of 147 common differential genes were screened from the two databases. Results of two-sample Mendelian randomization showed that GREM2 was associated with the increased risk of OSCC. In addition, SH3BGRL2 was associated with a decreased risk of OSCC, and DKK1, CCL11, and HOXC6 were considered as independent prognostic markers of OSCC. The predicted results of DKK1 were consistent with the actual results. KEGG enrichment analysis indicated the potential involvement of DKK1 in arachidonic acid and linoleic acid metabolism. Furthermore, DKK1 showed positive correlations with Tgd and Th2 cells, while displaying negative associations with PDC, Cytotoxic cells, Mast cells, CD8 T cells, TFH cells, B cells, T cells, and Th17 cells. Conclusion GREM2 is associated with an increased risk of OSCC. DKK1 is highly expressed in OSCC and associated with poor prognosis, which may be involved in regulating the metabolism of arachidonic acid and linoleic acid and immune cell invasion in OSCC.