Predicting clinical significance of BRCA1 and BRCA2 single nucleotide substitution variants with unknown clinical significance using probabilistic neural network and deep neural network-stacked autoencoder