Abstract |
Background: Modifiers significantly impact disease phenotypes by modulating the effects of disease-causing variants, resulting in varying disease manifestations among individuals. However, identifying genetic interactions between modifier and disease-causing variants is challenging. Results: We developed MDVarP, an ensemble model comprising 1000 random forest predictors, to identify modifier similar to disease-causing variant combinations. MDVarP achieves high accuracy and precision, as verified using an independent dataset with published evidence of genetic interactions. We identified 25 novel modifier similar to disease-causing variant combinations and obtained supporting evidence for these associations. MDVarP outputs a class label (Associated-pair or Nonrelevant-pair) and two prediction scores indicating the probability of a true association. Conclusions: MDVarP prioritizes variant pairs associated with phenotypic modulations, enabling more effective mapping of functional contributions from disease-causing and modifier variants. This framework interprets genetic interactions underlying phenotypic variations in human diseases, with potential applications in personalized medicine and disease prevention. |