SAV-Pred

SAV-Pred predicts the pathogenicity of single amino acid substitutions in proteins associated with monogenic hereditary diseases by using sequence-structure-property relationships to evaluate missense variants identified by Next Generation Sequencing (NGS), including those from newborn screening.

Key Features:

• Sequence-Structure-Property (SSPR) models: Represents protein fragments as molecular structural formulas to predict pathogenic effects of single amino acid substitutions.
• MultiPASS-derived models: SSPR models were developed from modifications to the MultiPASS cheminformatics software.
• NGS variant evaluation: Targets missense variants identified by Next Generation Sequencing (NGS) with emphasis on variants from newborn screening programs.
• Benchmarking against established predictors: Performance was compared with SIFT 4G, PolyPhen-2 HDIV, MutationAssessor, PROVEAN, and FATHMM.
• Performance metrics: Reported an average AUC of 0.804 ± 0.040 and higher quality scores for 15 of 25 proteins, with notably improved accuracy for IVD, HADHB, and HBB.
• Target disease set: Applied to twenty-five monogenic hereditary diseases included in the Uniform Screening Panel recommended by the Advisory Committee on Hereditary Disorders in Newborns and Children.

Scientific Applications:

• Variant interpretation in newborn screening: Supports interpretation of missense variants encountered in newborn screening and other NGS-based clinical workflows.
• Clinical and research diagnostics: Aids clinicians and researchers in classifying pathogenic amino acid substitutions in monogenic hereditary diseases.
• Benchmarking and tool comparison: Provides comparative benchmarking data against established predictors such as SIFT 4G, PolyPhen-2 HDIV, MutationAssessor, PROVEAN, and FATHMM.

Methodology:

Implements sequence-structure-property (SSPR) models representing protein fragments as molecular structural formulas; SSPR models were developed by modifications to MultiPASS and were benchmarked against SIFT 4G, PolyPhen-2 HDIV, MutationAssessor, PROVEAN, and FATHMM.