Utility of gene-specific algorithms for predicting pathogenicity of uncertain gene variants

The rapid advance of gene sequencing technologies has produced an unprecedented rate of discovery for genome variation in humans. A growing numbered of authoritative clinical repositories archive gene variants and disease phenotype, yet there are currently many more gene variants that lack clear annotation or disease association. To date, there has been very limited coverage of gene-specific predictors in the literature. Here we present the evaluation of ?gene-specific? predictor models based on a Na?ve Bayesian classifier for 20 gene-disease data sets, containing 3,986 variants with clinically characterized patient conditions. Utility of gene-specific prediction is then compared ?all-gene? generalized prediction and also to existing popular predictors. Gene-specific computational prediction models derived from clinically curated gene variant disease data sets often outperform established generalized algorithms for novel and uncertain gene variants.