Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Doctoral Award: Doctoral Foreign Study Award (DFSA)

Direct Payment

$105,000.00

Developing, Applying, and Explaining Graph Convolutional Neural Networks in the Context of Genome Wide Association Analyses to Increase Attention Deficit/Hyperactivity Disorder Prediction Accuracy, Extract Novel Genomic Networks Associated With Attention Deficit/Hyperactivity Disorder, and Improve Our Understanding of Attention Deficit/Hyperactivity Disorder Pathogenesis

FY 2021-22

Summary
Attention Deficit/Hyperactivity Disorder (ADHD) is a common psychiatric disorder that impairs patients' quality of life. Many factors, including genetics contribute to the disorder. Large scale studies of the genetics of people with and without ADHD have identified many small genetic changes linked to the disorder. By themselves, each genetic change is not helpful in predicting the likelihood that someone will develop ADHD; as a result, a new method which tabulates how many disease associated genetic changes a person has, yielding a 'risk score', has become popular. These scores have been helpful in predicting a person's ADHD status but are not sufficiently accurate for use in clinics. It is likely that individual genetic changes interact with one another in a more complicated way than these scores can capture. For instance, one group of changes might affect a certain biological function in some people, while another group of changes might do the same for others. This research project will use artificial intelligence to discover genes and gene networks that predict a person's ADHD status. The AI models will be "trained" by analyzing graphs, or pictorial representations of linked genetic changes, and learn what parts of a graph - or groups of changes - help decide if a person has ADHD or not. We will compare model accuracy to that of existing methods. We will then work backwards and figure out what networks the model used to make predictions. The first graphs we will use for prediction will be based on pre-existing networks, while the second set will discover new ones by looking at what genes combine to cause ADHD. Through this work, we will not only find out what genetic networks improve prediction accuracy, but will also improve our understanding of the neurobiology of ADHD. Our results will, eventually, lead to better treatments and methods to decide who, based on their genetics, is best suited for a particular treatment.
Program
Doctoral Award: Doctoral Foreign Study Award (DFSA)
Location
New York, United States of America
Type
Grant
Source
CIHR Funding Decisions Database

As published

Id
457223
Name
Saxena Ankita
Region
New York
Country
United States of America
Orgtype
Research Institution
Instname2
Neurosciences, Mental Health and Addiction
Theme2
Biomedical
Prcname2
Doctoral Research Awards - A
Projecttitle
Developing, Applying, and Explaining Graph Convolutional Neural Networks in the Context of Genome Wide Association Analyses to Increase Attention Deficit/Hyperactivity Disorder Prediction Accuracy, Extract Novel Genomic Networks Associated With Attention Deficit/Hyperactivity Disorder, and Improve Our Understanding of Attention Deficit/Hyperactivity Disorder Pathogenesis
Programtype2
Studentship Programs
Orgnameinp2
Direct Payment
Cihramount2
$105,000
Cihrequipment2
$0
Programname2
Doctoral Award: Doctoral Foreign Study Award (DFSA)
Keyworddelim
Adhd; Behavioral Disorders; Deep Learning; Genetics; Genomics; Machine Learning; Neural Network; Psychiatric Diseases
Abstract
Attention Deficit/Hyperactivity Disorder (ADHD) is a common psychiatric disorder that impairs patients' quality of life. Many factors, including genetics contribute to the disorder. Large scale studies of the genetics of people with and without ADHD have identified many small genetic changes linked to the disorder. By themselves, each genetic change is not helpful in predicting the likelihood that someone will develop ADHD; as a result, a new method which tabulates how many disease associated genetic changes a person has, yielding a 'risk score', has become popular. These scores have been helpful in predicting a person's ADHD status but are not sufficiently accurate for use in clinics. It is likely that individual genetic changes interact with one another in a more complicated way than these scores can capture. For instance, one group of changes might affect a certain biological function in some people, while another group of changes might do the same for others. This research project will use artificial intelligence to discover genes and gene networks that predict a person's ADHD status. The AI models will be "trained" by analyzing graphs, or pictorial representations of linked genetic changes, and learn what parts of a graph - or groups of changes - help decide if a person has ADHD or not. We will compare model accuracy to that of existing methods. We will then work backwards and figure out what networks the model used to make predictions. The first graphs we will use for prediction will be based on pre-existing networks, while the second set will discover new ones by looking at what genes combine to cause ADHD. Through this work, we will not only find out what genetic networks improve prediction accuracy, but will also improve our understanding of the neurobiology of ADHD. Our results will, eventually, lead to better treatments and methods to decide who, based on their genetics, is best suited for a particular treatment.
Competitiondate
202110
Approvedterm2
3 yrs 0 mth
Pinamesdelim
Saxena, Ankita
Approvedterm
3 ans 0 mois
Cihrcontribution2
~ $105,000~ $0~