When I analyze somatic mutations, I usually judge them according to many annotation filters such as conserved genomic regions, segmental_duplications and integrated databases (cosmic). Especially, I didn't believe any filter-based annotation in annovar which are SIFT or mutation_assessor because some cases have lower score but they are likely to be relevant to the cancer-disease or have dominant effects for disease.
From this points, I am wondering what useful annotation filter you use for mutation in exome seq?
Does any definite and reliable annotation filter exist?
It is always hard for bioinformaticians to differentiate artifacts or false_positives from real variants.