Abstract Number: PB0833
Meeting: ISTH 2022 Congress
Theme: Platelet Disorders, von Willebrand Disease and Thrombotic Microangiopathies » VWF and von Willebrand Factor Disorders - Clinical Conditions
Background: To determine if disease-causing variants (DCV) observed in patients with von Willebrand disease (VWD) are related to their clinical and laboratory phenotypes, additional costly, labor-intensive and time-consuming experimental approaches are needed. In-silico prediction tools were designed to predict the pathogenicity of DCV on the structure and/or function of the resulting protein. However, their performance can vary greatly.
A reliable statistical ratefor evaluating in-silico methods is to calculate the Matthews correlation coefficient (MCC) for each method, which considers true positives/negatives, false positives/negatives. Values≥0.7 indicate very strong agreement between prediction and observation.
Aims: • To predict pathogenicity of DCVs found in our VWD2 patients, using in-silico methods.
• To calculate MCC for each method in predicting both pathogenicity in DCVs and neutrality in benign single nucleotide variants (SNVs).
Methods: Thirty-one DCVs, 35 synonymous and 17 non-synonymousSNVs, all located within exons17-28of VWF gene.
Thirty in-silico methods: I-Mutant; PolyPhen; SIFT; SIFT4G; Mutation-Taster; Provean; DANN; CADD; FunSeq2; Predict-SNP2; GWAVA; Eigen; EigenPC; PhD-SNP; BayesDel-addAF; BayesDel-noAF; LRT; M-Cap; MVP; ListS2; MetaLR; MetaRNN; MetaSVM; MutPred; Revel; Deogen2; Mutation-Assessor; FATHMM-MKL; FATHMM-XF; FATHMM.
Results: All DCVs were predicted as pathogenic: 25/31(80.6%) by ≥70% of in-silico methods; 4/31 (12.9%), by < 40% of methods.
Synonymous SNVs were predicted as benign: 34/35(97.1%) by >70% of methods; 1/35(2.8%) by 57.1% of methods.
Non-synonymous SNVs were predicted as benign: 12/17(70.6%) by>70% of methods; 2/17(11.7%) by < 20 of methods; p.Pro1601Thr: pathogenic by all methods.
DANN, CADD, and FunSeq2 showed MCC≥0.7; MVP and I-mutant, the worst MCC
(Fig.1).
Conclusion(s): DANN, CADD, FunSeq2 and Predict-SNP2 showed the best performance. In-silico methods might be excellent tools for supporting the classification of DCVs related to VWD.
Synonymous SNVs showed higher predictive accuracy than non-synonymous SNVs. p.Pro1601Thr as benign variant should be revised.
Not all the in-silico methods discriminate between DCVs and SNVs. This point needs further analysis to improve them.
Fig. 1.
Matthews correlation coefficient values of in-silico methods
To cite this abstract in AMA style:
Woods A, Paiva J, Primrose D, Alberto M, Sanchez-Luceros A. Usefulness of in-silico prediction tools in the analysis of VWF genetic variants. [abstract]. https://abstracts.isth.org/abstract/usefulness-of-in-silico-prediction-tools-in-the-analysis-of-vwf-genetic-variants/. Accessed October 2, 2023.« Back to ISTH 2022 Congress
ISTH Congress Abstracts - https://abstracts.isth.org/abstract/usefulness-of-in-silico-prediction-tools-in-the-analysis-of-vwf-genetic-variants/