Unsupervised Representation Learning of DNA Sequences
This work addresses the challenge of handling variable-length DNA sequences in bioinformatics, though it is incremental as it applies an existing autoencoder method to a specific domain.
The authors tackled the problem of representing long, variable-length DNA sequences for classification tasks by using an unsupervised sequence-to-sequence autoencoder to learn fixed-dimension latent representations. They showed these representations can serve as features or priors for splice site classification, with qualitative analysis confirming identified splice signatures align with existing knowledge.
Recently several deep learning models have been used for DNA sequence based classification tasks. Often such tasks require long and variable length DNA sequences in the input. In this work, we use a sequence-to-sequence autoencoder model to learn a latent representation of a fixed dimension for long and variable length DNA sequences in an unsupervised manner. We evaluate both quantitatively and qualitatively the learned latent representation for a supervised task of splice site classification. The quantitative evaluation is done under two different settings. Our experiments show that these representations can be used as features or priors in closely related tasks such as splice site classification. Further, in our qualitative analysis, we use a model attribution technique Integrated Gradients to infer significant sequence signatures influencing the classification accuracy. We show the identified splice signatures resemble well with the existing knowledge.