Fig. 3.

Overview of the DeepSV. DeepSV extracts candidates training set from the original sequence, and then uses a gradient descent algorithm to train CNN network, and finally generates images according to gene sequence for calling deletions. The gradient descent algorithm is an optimization algorithm commonly used in machine learning and artificial intelligence to recursively approximate the minimum deviation model. a Finding candidate variants and Encoding pileup images. b Training CNN model. c Calling deletions