COVID-19 research needs interactive reliable biomarkers at genomic and genetic levels. Earlier research has established the existence of reliable interactive genomic biomarkers. However, reliable DNA methylation markers haven't been identified at the epigenetic level, not to mention interactiveness. This study, from 865,859 methylation sites, discovers two miniature sets of Infinium MethylationEPIC sites, each having seven (CpG sites, genes) to interact with each other and with disease subtypes. They lead to nearly perfect (96.87-100% accuracy) prediction of COVID-19 patients from patients with other diseases or healthy controls. These CpG sites can jointly explain some post-COVID-19-related conditions. These CpG sites and the optimum performed genomic biomarkers reported in our earlier work rise to be potential druggable targets. Among these CpG sites, diseases associated with cg16785077 (gene MX1) can have an incubation period of up to six to eight years, which raises a serious (or urgent) issue to investigate now or sooner. The gene PARP9 (cg25932713) is one of the only three human genes that contain both PARP domains and macrodomains. MX1 and PARP9 together were linked to SARS-CoV-1, MERS-CoV, and SARS-CoV-2. The new findings of CpG sites cg16785077 (gene MX1) and cg25932713 (PARP9) in COVID-19 at DNA methylation levels indicate that the initial SARS-CoV-2 may be better treated as transcribed viral DNA into RNA virus due to the long incubation feature of MX1 associated diseases, i.e., not as an RNA virus that has been concerned by scientists. Such a discovery will significantly change the scientific thinking and knowledge of viruses.