“Semi-Automated Human Genome Annotation” / “Deep Autoencoder Neural Networks for Gene Ontology Annotation Predictions”

“Semi-Automated Human Genome Annotation” by Michael M. Hoffman (principal investigator at the Princess Margaret Cancer Centre) Abstract: Sequence census methods like ChIP-seq now produce an unprecedented amount of genome-anchored data. We have developed an integrative method to identify patterns from multiple experiments simultaneously while taking full advantage of high-resolution data, discovering joint patterns across different assay types. We apply this method to ENCODE chromatin data for multiple human cell types, including ChIP-seq data on covalent histone modifications and transcription factor binding, and DNase-seq and FAIRE-seq readouts of open chromatin. In an unsupervised fashion, we identify patterns associated with transcription start sites, gene ends, enhancers, CTCF elements, and repressed regions. The method yields a model which elucidates the relationship between assay observations and functional elements in the genome. This model identifies sequences likely to affect transcription, and we verify these predictions in laboratory experiments. We have made software and an integrative genome browser track freely available (http://pmgenomics.ca/hoffmanlab/proj/segway/ ). “Deep Autoencoder Neural Networks for Gene Ontology Annotation Predictions” by Davide Chicco (postdoctoral fellow at the Princess Margaret Cancer Centre) Abstract: The annotation of genomic information is a major challenge in biology and bioinformatics. Existing databases of known gene functions are incomplete and prone to errors, and the biomolecular experiments needed to improve these databases are slow and costly. While computational methods are not a substitute for experimental verification, they can help in two ways: algorithms can aid in the curation of gene annotations by automatically suggesting inaccuracies, and they can predict previously-unidentified gene functions, accelerating the rate of gene function discovery. In this work, we develop an algorithm that achieves both goals using deep autoencoder neural networks. With experiments on gene annotation data from the Gene Ontology project, we show that deep autoencoder networks achieve better performance than other standard machine learning methods, including the popular truncated singular value decomposition.