Next-generation sequencing has provided extraordinary opportunities to investigate the massive human genetic variability. It helped identifying several kinds of genomic mismatches from the wild-type reference genome sequences and to explain the onset of several pathogenic phenotypes and diseases susceptibility. In this context, distinguishing pathogenic from functionally neutral amino acid changes turns out to be a task as useful as complex, expensive, and time-consuming. Here, we present an exhaustive and up-to-dated survey of the algorithms and software packages conceived for the estimation of the putative pathogenicity of mutations, along with a description of the most popular mutation datasets that these tools used as training sets. Finally, we present and describe software for the prediction of cancer-related mutations.