method implemented in a neural network of training of a splice site detector that identifies splice sites in genomic sequences, trained splice site predictor and system

摘要

the technology disclosed refers to the construction of a classifier based on convolutional neural network for the classification of variants. in particular, it refers to training a classifier based on convolutional neural network in training data using a gradient update technique based on backpropagation that progressively combines the outputs of the classifier based on convolutional neural network with corresponding ground truth markers. the classifier based on convolutional neural network comprises groups of residual blocks, each group of residual blocks is parameterized by a number of convolution filters in the residual blocks, a size of the convolution window of the residual blocks and an atrous convolution rate of the residual blocks, the size of the convolution window varies between groups of residual blocks, the atrous convolution rate varies between groups of residual blocks. the training data includes benign training examples and pathogenic training examples of translated sequence pairs generated from benign and pathogenic variants.