Frequent pathological human mutations: a survey

We analyzed some mutations from dbSNP database which were known to be pathological, but are relatively frequent in humans with the intention of testing PolyPhen-2's efficiency on particularly difficult cases. We proved that the majority of these mutations could benefit from a therapeutic approach with pharmacological chaperones. Exon-sequencing will soon become a common practice in disease diagnoses. All the steps that follow the acquisition of the sequence must be optimized if we wish to find one or more mutations associated to the pathological phenotype. We believe the diagnosis will expand from only working on early onset and severe phenotype diseases to working on late onset and mild phenotype diseases too. Therefore the cases of research on mutations that cause little damage to the mutant protein will get more and more frequent. One of the most common programs used to distinguish pathological mutations is PolyPhen-2[1]. PolyPhen-2 algorithm classifies variations using eight sequence-based and three structure-based predictive features. We analyzed cases that may be found in clinical practice with a relative frequency (minor allele frequency MAF>0). We found that in half of the cases PolyPhen-2 erroneously classifies these mutations as benign or neutral. The percentage of false negatives does not decrease when the 3D structure of the protein is known and PolyPhen-2 can also use three structure-based predictive features. The prediction becomes more precise when