I was checking some of the variants I have found in cancer samples that I have been analysing. I am not much aware of how COSMIC data should be interpreted. So if something is already found in COSMIC, can we consider that variant novel or something significantly important for publication?
COSMIC does provide a large catalogue of somatic variants from many sequencing studies, although their data can be somewhat messy (they include variants sometimes which are not marked as somatic). Examining COSMIC can be reasonably good evidence at determining whether somatic variants have been seen before. That said, the majority of somatic variants in COSMIC are passenger mutations, and even if a hypothetically true driver mutation has been seen in COSMIC a few times does not mean people have previously implicated it as a driver (you might have to backtrack the study to examine their conclusions though). In fact, rare driver genes (especially <5-10% of samples) are generally poorly characterized in most tumor types (http://www.nature.com/nature/journal/v505/n7484/abs/nature12912.html). COSMIC updates regularly, but the current version may not have included all studies recently performed. So I would suggest that a literature search for very recent studies related to your particular cancer samples would still be warranted. Lastly and perhaps most important, whether your variants are "important" greatly matters on whether they are driver or passenger mutations. A passenger mutation that has not been seen before is not really important. Whether mutations may be drivers can be supported statistically/computationally and/or by experiments. Some support for driver status may be whether your cancer samples support the gene as significantly mutated (e.g. MutsigCV, MUSIC), high in silico predicted score for mutations (e.g. CHASM, VEST), and for protein coding genes where the mutation occurs in 3D protein structure (MUPIT maps cancer mutations to 3D protein structure with functional annotations).