I am dealing with a very special problem: I have to perform DESeq on the count data that belong to a single gene (only ONE gene, i.e. the count data has one row & 100 columns). The dispersion value is already available for the gene. (Actually, the dispersions were calculated by DESeq over a master count data with 19k genes )

Question: How can I use DESeq on this special count data?! I have to somehow feed the DESeq with the already-known dispersion value, becauseDESeq cannot calculate dispersion when there's only one gene in the count data.

You can just use dispersions(dds) <- ... to supply your own dispersion estimates rather than estimating them with DESeq2. Instead of running DESeq() you would run the individual steps, see ?DESeq.