Web-Based Bioinformatics Predictors: Recommendations to Assess Lysosomal Cholesterol Trafficking Diseases-Related Genes

Introduction The growing number of genetic variants of unknown significance (VUS) and availability of several in silico prediction tools make the evaluation of potentially deleterious gene variants challenging. Materials and Methods We evaluated several programs and software to determine the one that can predict the impact of genetic variants found in lysosomal storage disorders (LSDs) caused by defects in cholesterol trafficking best. We evaluated the sensitivity, specificity, accuracy, precision, and Matthew's correlation coefficient of the most common software. Results Our findings showed that for exonic variants, only MutPred1 reached 100% accuracy and generated the best predictions (sensitivity and accuracy=1.00), whereas intronic variants, SROOGLE or Human Splicing Finder (HSF) generated the best predictions (sensitivity=1.00, and accuracy=1.00). Discussion Next-generation sequencing substantially increased the number of detected genetic variants, most of which were considered to be VUS, creating a need for accurate pathogenicity prediction. The focus of the present study is the importance of accurately predicting LSDs, with majority of previously unreported specific mutations. Conclusion We found that the best prediction tool for the NPC1 , NPC2 , and LIPA variants was MutPred1 for exonic regions and HSF and SROOGLE for intronic regions and splice sites.