A model averaging approach for improved in silico variant prediction
A variety of in silico scores are currently utilized in missense variant assessment as evidence of pathogenicity. Information from these scores is typically selected manually or using stepwise regression procedures.
We developed a gene-specific model averaging approach that improved model robustness and prediction performance in assigning benign and pathogenic evidence to missense variants when assessing a collection of in silico scores.
Our model averaging method also provides a gene-specific relative importance index for each in silico score, which may be a valuable resource for variant interpretation.
Authors: Yuan Tian; Shuwei Li; Mary Helen Black; Aaron Elliott; Dajun Qian