The SOPHiA Platform normalizes the scores provided by the different databases to enable comparisons among the difference sources. For all predictive scores, a value of 1 means likely pathogenic and 0 likely benign irrespective of the database. The scores are transformed as:
SIFT
“1 – SIFT” score (the original SIFT score is opposite, with 0 being predicted pathogenic and 1 benign)
PolyPhen2
No transformation (there are two different PolyPhen2 scores, HVAR and HDIV, based on different training sets)
MutationTaster
Combination of prediction and score (1 – confident, 0 not confident). The SOPHiA Platform displays the score if the prediction is pathogenic or “1 – score” if the prediction is benign
LRT score (not in spider plot but can be selected in the variant table)
Contains the score and a value omega. The SOPHiA Platform displays “1 – LRT score x 0.5” if omega < 1 or “LRT score x 0.5” if omega ? 1
More detail of such transformations can be found in the dbNSFP database publication: