Course Number: 02-718
Modern medical research increasingly relies on the analysis of large patient datasets to enhance our understanding of human diseases. This course will focus on the computational problems that arise from studies of human diseases and the translation of research to the bedside to improve human health. The topics to be covered include computational strategies for advancing personalized medicine, pharmacogenomics for predicting individual drug responses, metagenomics for learning the role of the microbiome in human health, mining electronic medical records to identify disease phenotypes, and case studies in complex human diseases such as cancer and asthma. We will discuss how machine learning methodologies such as regression, classification, clustering, semi-supervised learning, probabilistic modeling, and time-series modeling are being used to analyze a variety of datasets collected by clinicians. Class sessions will consist of lectures, discussions of papers from the literature, and guest presentations by clinicians and other domain experts. Grading will be based on presentations, assignments, participation, and a project.
Prerequisite(s): The course is designed for graduate and upper-level undergraduate students with a wide variety of backgrounds. Students should have some background in Machine Learning, but no prior background in Medicine is required.
- Homeworks (60%)
- Five homeworks will be assigned.
- Every student is given a budget of three 'late days' to use as they see fit.
- There is no need to email the instructor to request a late day
- Anything that is more than 15 minutes past the deadline is considered late.
- If you hand the assignment more than 24 hrs after the deadline, two late days will be changed, etc.
- Once you use up your three late days, a 25% per day penalty is applied to the homework grade.
- Course Project (40%)
- You will complete a data analysis project and write a short report and give a short (10-15 min) presentation.
- A project proposal will be due mid-semester. See Proposals for more information.
- Cheating policy: All work must be your own and novel. Unauthorized collaboration, falsified data, or plagiarism will result in a failing grade and will be reported to your academic advisor and dean.
- Double dipping policy: You may not re-use data, reports, manuscripts, or publications from your research or from other courses. However, you may extend your previous work, as long as you inform the instructor that you are doing so. Please contact the instructor if you have any questions regarding this policy.