I mention a couple [for somatic mutations] at the bottom, here: A: pathogenicity predictors of cancer mutations

Kevin

You could check out OpenCRAVAT, which has many annotations related to somatic mutations in cancer. This includes quite a few of the machine learning predictions for driver mutations in cancer, which Kevin pointed out in his post. I personally recommend CHASMplus, but I am a little bit biased since I'm the developer of CHASMplus. OpenCRAVAT also provides variant annotations from databases such as CIViC, which manually curates information on somatic mutations. Other cool features are that you can visualize your variants on available protein structures via MuPIT.