PON-P3: Accurate Prediction of Pathogenicity of Amino Acid Substitutions DOI Open Access
Muhammad Kabir, Saeed Ahmed, Haoyang Zhang

и другие.

International Journal of Molecular Sciences, Год журнала: 2025, Номер 26(5), С. 2004 - 2004

Опубликована: Фев. 25, 2025

Different types of information are combined during variation interpretation. Computational predictors, most often pathogenicity provide one type for this purpose. These tools based on various kinds algorithms. Although the American College Genetics and Association Molecular Pathology guidelines classify variants into five categories, practically all predictors binary pathogenic/benign predictions. We developed a novel artificial intelligence-based tool, PON-P3, basis carefully selected training dataset, meticulous feature selection, optimization. started with 1526 features describing variations, their sequence structural context, parameters affected genes proteins. The final random boosting method was tested compared total 23 predictors. PON-P3 performed better than recently introduced which utilize large language models or methods that use evolutionary data alone in combination different gene protein properties. classifies cases three categories as benign, pathogenic, uncertain significance (VUSs). When test were used, some metapredictors slightly PON-P3; however, real-life situations, patient data, those overpredict both pathogenic benign cases. predicted possible amino acid substitutions human proteins encoded from MANE transcripts. also used to predict unambiguous VUSs (i.e., without conflicts) ClinVar. A 12.9% be 49.9% benign.

Язык: Английский

PON-P3: Accurate Prediction of Pathogenicity of Amino Acid Substitutions DOI Open Access
Muhammad Kabir, Saeed Ahmed, Haoyang Zhang

и другие.

International Journal of Molecular Sciences, Год журнала: 2025, Номер 26(5), С. 2004 - 2004

Опубликована: Фев. 25, 2025

Different types of information are combined during variation interpretation. Computational predictors, most often pathogenicity provide one type for this purpose. These tools based on various kinds algorithms. Although the American College Genetics and Association Molecular Pathology guidelines classify variants into five categories, practically all predictors binary pathogenic/benign predictions. We developed a novel artificial intelligence-based tool, PON-P3, basis carefully selected training dataset, meticulous feature selection, optimization. started with 1526 features describing variations, their sequence structural context, parameters affected genes proteins. The final random boosting method was tested compared total 23 predictors. PON-P3 performed better than recently introduced which utilize large language models or methods that use evolutionary data alone in combination different gene protein properties. classifies cases three categories as benign, pathogenic, uncertain significance (VUSs). When test were used, some metapredictors slightly PON-P3; however, real-life situations, patient data, those overpredict both pathogenic benign cases. predicted possible amino acid substitutions human proteins encoded from MANE transcripts. also used to predict unambiguous VUSs (i.e., without conflicts) ClinVar. A 12.9% be 49.9% benign.

Язык: Английский

Процитировано

0