In silico prediction of deleteriousness for nonsynonymous and splice-altering single nucleotide variants in the human genome

In silico prediction methods have increasingly been valuable and popular in molecular biology, especially in human genetics, for deleteriousness prediction to filter and prioritize huge amounts of DNA variation identified by sequencing human genomes. There is a rich collection of available methods developed upon different levels/aspects of knowledge about how DNA variations affect gene expression. Given the fact that their predictions are not always consistent or even opposite of what was expected, using consensus prediction or majority vote among these methods is preferred to trusting any single one. Because querying different databases for different methods is both tedious and time-consuming for such big data sets, one database integrating predictions from multiple databases can facilitate the process. In this chapter, we describe the general steps of obtaining comprehensive predictions and annotations for large numbers of variants from dbNSFP, the first and probably the most widely used database of its kind.