Workshop in Biostatistics

Medical School Office Building (MSOB)
Rm x303

DATE: November 30, 2017
TIME: 1:30 - 2:50 pm
TITLE: Statistical learning approaches to understand the genetic and phenotypic complexity of autism
SPEAKER:
Dennis P. Wall
Associate Professor of Pediatrics (Systems Medicine), of Biomedical Data Science and, by courtesy,
of Psychiatry and Behavioral Sciences, Stanford

 

Abstract:
I will discuss my lab’s efforts to use machine learning to build practical models for better autism care and to tie genotype to phenotype. I will begin by defining autism and why & how it represents a quintessential translational bioinformatics challenge. Then, I will describe our use of machine learning [1], including sparsifying models to find top ranked features that align well clinical diagnosis but that can be used in mobile detection efforts. Next, I will discuss our use of generalized low rank models [2] to impute missingness and improve our ability to define the forms of autism. Finally, I will describe approaches to recode the genome of autism subjects into vectors that can be used for prioritizing genes via coalitional game theory [3] and that can also be used for predicting phenotype(s).

Suggested readings:

[1] Kosmicki, J., Sochat, V., & Wall, D.P. (2015). Searching for a minimal set of behaviors for autism detection through feature selection based machine learning. Nature Translational Psychiatry, v.5(2): e514.  doi:10.1038/tp.2015.7.

[2] Udell M, Horn C, Zadeh R, Boyd S. Generalized low rank models. Foundations and Trends in Machine Learning. 2016 Jun 23; 9(1):1-118.

[3] Esteban, F. & Wall D.P. (2009). Using game theory to detect genes involved in Autism Spectrum Disorder. TOP. July 2011, Vol. 19, Issue 1, pp 121-129.  (ISSN 1863-8279).