Alipanahi, B. et al. Nat. Biotechnol. 33, 831–838 (2015).
To understand the role of DNA- and RNA-binding proteins in gene regulation, it is essential to characterize their sequence specificities. Experimental evidence for sequence specificity comes in many different formats, including the output from protein binding microarrays, chromatin immunoprecipitation and high-throughput SELEX (systematic evolution of ligands by exponential enrichment). To analyze such disparate data types in very large data sets, Alipanahi et al. introduce DeepBind, a software tool based on deep convolutional neural networks. Sequence specificities can be visualized as a 'mutation map' that shows the effect of variations on binding. The authors applied DeepBind to study the role of RNA-binding proteins in regulating alternative splicing and to analyze how disease-associated variants affect transcription factor binding and gene expression.
Rights and permissions
About this article
Cite this article
Deep learning to predict sequence specificity. Nat Methods 12, 809 (2015). https://doi.org/10.1038/nmeth.3559
Published:
Issue Date:
DOI: https://doi.org/10.1038/nmeth.3559