Performance evaluation of computational methods for splice-disrupting variants and improving the performance using the machine learning-based framework.