SemanticCAP: Chromatin Accessibility Prediction Enhanced by Features Learning from a Language Model (2204.02130v2)

Abstract: A large number of inorganic and organic compounds are able to bind DNA and form complexes, among which drug-related molecules are important. Chromatin accessibility changes not only directly affects drug-DNA interactions, but also promote or inhibit the expression of critical genes associated with drug resistance by affecting the DNA binding capacity of TFs and transcriptional regulators. However, Biological experimental techniques for measuring it are expensive and time consuming. In recent years, several kinds of computational methods have been proposed to identify accessible regions of the genome. Existing computational models mostly ignore the contextual information of bases in gene sequences. To address these issues, we proposed a new solution named SemanticCAP. It introduces a gene LLM which models the context of gene sequences, thus being able to provide an effective representation of a certain site in gene sequences. Basically, we merge the features provided by the gene LLM into our chromatin accessibility model. During the process, we designed some methods to make feature fusion smoother. Compared with other systems under public benchmarks, our model proved to have better performance.