Dilated Convolutions for Modeling Long-Distance Genomic Dependencies
This addresses the challenge of modeling long-distance genomic dependencies for bioinformatics researchers, though it appears incremental as it applies an existing method (dilated convolutions) to a new genomic dataset.
The researchers tackled the problem of detecting regulatory elements in the human genome by modeling long-distance dependencies, which are difficult with small DNA snippets. They showed that dilated convolutions effectively locate regulatory markers like transcription factor binding sites, histone modifications, and DNAse hypersensitivity sites.
We consider the task of detecting regulatory elements in the human genome directly from raw DNA. Past work has focused on small snippets of DNA, making it difficult to model long-distance dependencies that arise from DNA's 3-dimensional conformation. In order to study long-distance dependencies, we develop and release a novel dataset for a larger-context modeling task. Using this new data set we model long-distance interactions using dilated convolutional neural networks, and compare them to standard convolutions and recurrent neural networks. We show that dilated convolutions are effective at modeling the locations of regulatory markers in the human genome, such as transcription factor binding sites, histone modifications, and DNAse hypersensitivity sites.