Table 2.

A summary of functional prediction scores and conservation scores.

Score	Training data	Information used	Prediction model
PolyPhen-2	UniProtKB/UniRef100; PDB/DSSP; UCSC alignments of 45 vertebrate genomes	eight sequence-based and three structure-based predictive features	naive Bayes classifier
SIFT	SWISS-PROT/TrEMBL	sequence homology based on PSI-BLAST	position specific scoring matrix
Mutation Taster	UniProt; homologous genes in humans and 10 other species; dbSNP; HapMap	conservation, splice site, mRNA features, protein features;	naive Bayes classifier
LRT	coding sequences of 32 vertebrate species	sequence homology	likelihood ratio test of codon neutrality
Mutation Assessor	homologous sequences from Uniprot identified by BLAST	sequence homology of protein families and sub-families within and between species	combinatorial entropy formalism
FATHMM	homologous sequences from UniRef90, SUPERFAMILY and Pfam	sequence homology	hidden Markov models
SiPhy	genomes of 29 mammals	multiple alignments	inferring nucleotide substitution pattern per site
GERP++	genomes of 34 mammals	multiple alignments and phylogenetic tree	maximum likelihood evolutionary rate estimation
PhyloP	genomes of 33 placental mammals	multiple alignments and phylogenetic tree	distributions of the number of substitutions based on phylogenetic hidden Markov model