Personalized Medicine via Network Analysis of ABCD Study

Mentors:
Lead: Nathaniel Josephs (Statistics, NCSU)
Collaborator: Shannon Ford (Health and Human Services, UNC Wilmington)

Intellectual merit and significance:
The Adolescent Brain Cognitive Development (ABCD) study is the largest study of brain and health development in US children [1]. The results of the ABCD Study have the potential to promote the well-being of our future generations, but actionable information needs to be provided to health professionals. Currently, the volume of social and behavioral data makes creating simple recommendations difficult. Further complicating the analysis is the longitudinal nature of the study, but identifying cognitive or other health changes over time may be precisely a key that unlocks personalized medicine recommendations.

Objectives:
This study proposes to use a network analysis [2] to disentangle the complex dependencies in brain and health development. We will begin by constructing demographic-disease networks that can be used to visualize the complex relationships that govern our health. We will then look for network predictors that correlate with various health outcomes [3]. Afterwards, we will perform a statistical analysis on dynamic networks to try to understand if changes in the network topologies relate to changes in health outcomes [4]. Time-permitting, we will perform an observational causal analysis of the data using a case-control study design.

Outcomes:
Students will develop hands-on training with networks using the R package igraph. They will learn to construct, visualize, and perform EDA on networks. Students will also learn how to fit statistical models with network inputs, along with the tenets of causal analysis.

References:
[1] https://abcdstudy.org
[2] Kolaczyk, E.D. and Csárdi, G., 2014. Statistical analysis of network data with R (Vol. 65). New York: Springer.
[3] Rudolph, A.E., Upton, E., Young, A.M. and Havens, J.R., 2021. Social network predictors of recent and sustained injection drug use cessation: findings from a longitudinal cohort study. Addiction, 116(4), pp.856-864.
[4] Krivitsky, P.N. and Handcock, M.S., 2014. A separable model for dynamic networks. Journal of the Royal Statistical Society Series B: Statistical Methodology, 76(1), pp.29-46.