Ensemble and consensus approaches to prediction of recessive inheritance for missense variants in human disease

Area of Science:

• Genetics and Genomics
• Bioinformatics
• Clinical Diagnostics

Background:

• Mode of inheritance (MOI) is crucial for interpreting pathogenic genetic variants, yet this information is often missing.
• Existing variant effect prediction tools struggle with identifying variants causing recessive-acting diseases.
• Accurate MOI determination is essential for precise clinical genetic variant interpretation.

Purpose of the Study:

• To develop computational tools that predict variant pathogenicity while accounting for mode of inheritance.
• To create a consensus method integrating multiple MOI prediction tools for improved accuracy.
• To validate the clinical utility of these tools using real-world electronic health record data.

Main Methods:

• Developed MOI-Pred, a tool integrating evolutionary and functional annotations for variant-level pathogenicity predictions.
• Created ConMOI, a consensus method combining predictions from three independent MOI prediction tools.
• Validated predictions using a large-scale electronic health record (EHR) dataset of 29,981 individuals.

Main Results:

• MOI-Pred and ConMOI demonstrate state-of-the-art performance on standard benchmarks.
• Predictions from both tools showed significant enrichment for pathogenic variants in dominant and recessive diseases.
• ConMOI outperformed its individual component methods in both benchmarking and EHR-based validation.

Conclusions:

• MOI-Pred and ConMOI effectively predict pathogenicity for both dominant and recessive variants.
• The consensus approach (ConMOI) enhances prediction accuracy and robustness.
• These tools represent a significant advancement in variant interpretation, aiding clinical genetic diagnostics.