DiGePred

DiGePred predicts candidate digenic disease gene pairs to prioritize gene-pair combinations that may underlie rare genetic disorders.

Key Features:

• Random forest classifier: Uses a random forest classifier to predict digenic disease-causing gene pairs.
• Multimodal feature set: Leverages features derived from biological networks, genomics, evolutionary history, and functional annotations.
• Training data: Trained on DIDA (the largest available database of known digenic-disease-causing gene pairs) and multiple non-digenic gene-pair sets, including variant pairs from unaffected relatives of UDN individuals.
• Evaluation: Validated by cross-validation and a held-out test set with a precision-recall area under the curve exceeding 77%.
• False-positive control: Designed to control false positives effectively in realistic clinical settings.
• Genome-wide predictions: Provides precomputed predictions for all human gene pairs.

Scientific Applications:

• Clinical prioritization: Prioritizes candidate digenic gene pairs for diagnosing rare genetic disorders.
• Variant-pair screening: Enables rapid screening of variant gene pairs in research and clinical genomics workflows.
• Discovery of non-monogenic causes: Facilitates discovery of genetic causes for rare non-monogenic diseases.

Methodology:

Employs a random forest classifier trained on DIDA and non-digenic gene-pair sets (including UDN unaffected-relative variants) using features from biological networks, genomics, evolutionary history, and functional annotations, and evaluated by cross-validation and a held-out test set.