Clustering in single cell data utilizes principal components rather than individual genes. So to do what you want, you'd just use those genes for your PCA rather than the top X most variable ones or whatever. In Seurat, you can just feed your genes of interest to the features parameter of the RunPCA function prior to clustering as usual.

Note that I can't say I'd recommend this, as it'd ignore any other potential differences or similarities between cells and may cluster things together that really aren't all that biologically similar depending on whatever you feed in.