I am planning to try dbNSFP in order to get consistent and comparable scores of protein mutation consequences. I noticed that VEP outputs four types of SIFT scores ("Deleterious", "Deleterious-low confidence", "Tolerated", "Toleraged-low confidence"). I wonder whether/how dbNSFP does take into account the "low confidence" version of these scores ? Are those ignored or integrated in different ways with respect to "Deleterious" and "Tolerated" ? I am also looking to better understand this SIFT low-confidence labelling of the scores...

Thx a lot