What genes cause cancer ? Have we inherited genes from Neanderthals ? How does a single genome code for the diverse function that we see?
We can now begin to answer these fascinating questions in biology because the cost of genome sequencing has fallen faster than Moore's law. The bottleneck in answering these questions has shifted from data generation to powerful statistical models and inference algorithms that can make sense of this data. Statistical machine learning provides an important toolkit in this endeavor. Further, biological datasets offer new challenges to the field of machine learning.
We will learn about probabilistic models, inference and learning in these models, model assessment, and interpreting the inferences to address the biological questions of interest. The course aims to introduce CS/Statistics students to an important set of problems and Bioinformatics/Human Genetics students to a rich set of tools.
Familiarity with probability, statistics, linear algebra and algorithms is expected. No familiarity with biology is needed.
Instructor: Sriram Sankararaman
Office Hours: Boelter 4531D, Wednesday 1:00  2:00p (or by appointment)
Email: sriram at cs dot ucla dot edu
You are free to discuss the homework problems. However, you must write up your own solutions. You must also acknowledge all collaborators.
The course website is based on material developed by Ameet Talwalkar and Fei Sha. Some of the administrative content on the course website is adapted from material from Jenn Wortman Vaughan, Rich Korf, and Alexander Sherstov.
Date  Topics  Reading  HW 

09/26  Introduction to genomics  Big Data: Astronomical or Genomical? Nova: Personal DNA testing 

09/28  
10/03  Association studies: linear regression 
Homework 1 

10/05  Association studies: logistic regression  
10/10  Heritability: ridge regression and mixed models  
10/12  Clustering and mixture models  
10/17  The EM algorithm 
Homework 2 Data for Homework 2 

10/19  No class  
10/26  PCA  
10/28  PCA and probabilistic PCA  
10/31  Admixture models 
Homework 3 Data for Homework 3 

11/02  Applications of admixture models and population stratification  
11/07  Directed graphical models (DGMs)  
11/09  DGMs: Conditional independence  
11/14  DGMs. Hidden Markov Models  
11/16  Midterm  
11/21  Hidden Markov Models 
Homework 4+5 Data for Homework 4+5 

11/23  Kernels  
11/28  Neural networks and deep learning  
11/30  Genomic privacy 