23sep12:00 pm1:00 pmMachine learning approaches to discover biological regulatory features from high-throughput sequencing data12:00 pm - 1:00 pm BTM 2006B Lecturer: Qiwen Hu Event Category:Invited talk,Statistics and machine learning
Machine learning involves the creation and evaluation of algorithms related to classification, prediction and pattern recognition based on modeling a set of observed data. Machine learning has been successfully applied
Machine learning involves the creation and evaluation of algorithms related to classification, prediction and pattern recognition based on modeling a set of observed data. Machine learning has been successfully applied in many areas such as image recognition and natural language processing. The rapid expansion of publicly available biological datasets provides great opportunities for discovery-driven research with poorly understood mechanisms and diseases. In this talk, I will introduce how machine learning models canbe used to extract biological meaningful regulatory signals from high-throughput sequencing data. We developed machine learning models based on different strategies in order to identify regulatory elements that contribute to transcription and translation in different model systems as well as disease-related signals from large publicly available datasets. Using supervised, semi-supervised and unsupervised approaches, we found that specific epigenetic factors such as histone marks are associated with and predict gene splicing patterns in the nucleus accumbens – a brain reward region related to drug addiction and developmental tissues in mammalian, with particular histone marks showing the significant enrichment at alternative spliced exons. We also identified the translated upstream open reading frames – a small sequence elements located upstream of a functional gene that may block the translation of main coding regions of the entire gene and the limitation of current models to analyze high dimensional sequencing datasets. Our results will provide the basis for predictive models of transcription and translation and contribute to a better understanding of regulatory mechanism in different model organisms.
(Monday) 12:00 pm - 1:00 pm
60 Fenwood Road, Boston, MA 02115