We used the logistic regression-based deleterious mutation prediction data of dbNSFP, which includes the scores of ten different tools (SIFT, PolyPhen-2 HDIV, PolyPhen-2 HVAR, GERP++, Mutation Taster MutationAssessor, FATHMM, LRT, SiPhy, and PhyloP).
DG, SH, MD, TM, and GR analyzed the somatic mutation prediction and validation data.
This study assessed the performance of mutation prediction models in this population-based data set.
We have presented here an easily implemented but highly effective strategy for enriching true somatic SNVs in exome-sequencing data, using consensus between multiple somatic mutation prediction algorithms.
Thus, mutation prediction software was ultimately inadequate at predicting disease class, as mutations from both class 1 and class 2 were found with similar PROVEAN scores.
The workflow of our PSAAP algorithm is summarized in Figure 3. Next, both curated RET mutations (known disease association) and RET uncertain variants (VUS data) were analyzed and compared using four existing mutation prediction algorithms.
The data sets used in this work can be divided into four groups by predictive task: prediction of protein stability change on mutation, prediction of protein protein and protein DNA affinity change on mutation.
Schwarz JM, Cooper DN, Schuelke M, Seelow D. MutationTaster2: mutation prediction for the deep-sequencing age.
Of the tested mutation prediction software, not one was able to consistently distinguish between class 1 and class 2 mutations.
Schwarz, J. M., Cooper, D. N., Schuelke, M. & Seelow, D. MutationTaster2: mutation prediction for the deep-sequencing age.
The three aforementioned mutation prediction software classified 100% of class 2 mutations as probably damaging (Supplementary Table 1).
