Machine Learning for Genomics

Catalog description

This course focuses on how a variety of topics in machine learning, deep learning, and statistical models are applied in current genomic research that involve large scale sequencing data such as DNA/RNA sequencing, or RNA/DNA protein binding assays. DL topics include transformers, CNN, (V)AE, and interpretation methods. ML and statistical modeling include variants of clustering and regression, MCMC, ensemble learning, transfer learning, dimensionality reduction, dealing with missing data, imbalanced data, measures of performance. All of those will be covered within the context of current research in areas such as multi-omics integration, single cell transcriptomics, cancer genomics, RNA processing, Genome wide association studies, and genetic variant fine mapping. A major focus is how to formulate a scientific question, a model/algorithm that could answer the question, and how to then assess what the model/algorithm produces. Students will read and critically discuss seminal papers in applied machine/deep learning for genomic research, perform “realistic” genomic research tasks (code+analysis), leading to a final project where they would have to formulate a question in genomic research that would require them to develop and apply methods to answer it. Recommended preparation for the course would be CIS 5190 /5200 or a graduate level machine learning course.