CSE428: Computational Biology Capstone

Instructor: Sara Mostafavi, PhD. saramos@cs

Teaching Assistant: Ayse Dincer. abdincer@cs

Time and place: Tuesday / Thursday 10:00-11:20am, LOW 112

Office hours: By appointment

Discussion Board: Ed

Description:

Recent advances in genomic measurement technologies now enable scientists to generate different types of high-dimensional data, including data from our genomes, epigenomes and transcriptomes, resulting in millions of measurements per individual. These data promise to enable prediction of disease risk and understand its molecular causes. The field of computational biology brings together recent advances in computational methods, including statistical reasoning and machine learning, and genomic measurement technologies, to provide insights into how biological systems, like a cell, work and what molecular events/mechanisms lead to health vs. disease.

The primary objective of the Computational Biology Capstone is to give CSE students an integrative research experience in this highly interdisciplinary field. Working in small research teams, the students will learn how to apply state-of-the-art machine learning models to make rigorous predictions from various types of genomics data.

Specifically, this quarter, we will focus on a set of related projects that involve applying deep neural networks (i.e., deep learning) to make predictions using genomic sequence data as input, referred to as "Seq2Function" models. Through the course of the projects, students will learn 1) how to read and critically assess the latest research papers in this field, 2) train deep learning models on complex and heterogeneous datasets, 3) critically assess model performance, 4) derive an understanding of learned features and biological mechanisms using explainable AI techniques.

Learning objectives

• Exposure to cutting-edge research directions in Computational Biology

• Learn to apply the latest deep learning models to genomic datasets

• Learn to apply emerging explainable AI techniques to derive and understand features learned by complex deep learning models

• Identify strategies to address research challenges, including design of baseline experiments, and model evaluation

Evaluation:

Class discussion, progress reports, final oral and written project report, peer evaluation by teammates.

See Schedule page for weekly topics

See Project page for a description of potential projects