The disclosed technique presents a deep learning-based framework that identifies sequence patterns that cause sequence-specific errors (SSEs). Systems and methods train variant filters on large variant data to learn the causal dependencies between sequence patterns and false variant calls. Variant filters have a hierarchical structure built on deep neural networks such as convolutional neural networks and fully coupled neural networks. Systems and methods use variant filters to perform simulations that test for known sequence patterns for their effect on variant filtering. The premise of the simulation is as follows. When a pair of tested repeat pattern and called variant is fed to the variant filter as part of the simulated input array and the variant filter classifies the called variant as an incorrect variant call, then It is possible that the iterative pattern was identified as causing a false variant call and causing SSE.
展开▼
