Estimating a brain network predictive of stress and genotype with supervised autoencoders

Austin Talbot; David Dunson; Kafui Dzirasa; David Carlson

doi:10.1093/jrsssc/qlad035

Estimating a brain network predictive of stress and genotype with supervised autoencoders

J R Stat Soc Ser C Appl Stat. 2023 May 22;72(4):912-936. doi: 10.1093/jrsssc/qlad035. eCollection 2023 Aug.

Authors

Austin Talbot¹, David Dunson², Kafui Dzirasa^{3

4

5

6

7}, David Carlson^{8

9

10}

Affiliations

¹ Pillar Biosciences Inc., Natick, MA, USA.
² Department of Statistical Science, Duke University, Durham, NC, USA.
³ Department of Psychiatry and Behavioral Sciences, Duke University, Durham, NC, USA.
⁴ Department of Neurobiology, Duke University, Durham, NC, USA.
⁵ Department of Neurosurgery, Duke University, Durham, NC, USA.
⁶ Department of Biomedical Engineering, Duke University, Durham, NC, USA.
⁷ Howard Hughes Medical Institute, Chevy Chase, MD, USA.
⁸ Department of Civil and Environmental Engineering, Duke University, Durham, NC, USA.
⁹ Department of Biostatistics and Bioinformatics, Duke University, Durham, NC, USA.
¹⁰ Department of Computer Science, Duke University, Durham, NC, USA.

Abstract

Targeted brain stimulation has the potential to treat mental illnesses. We develop an approach to help design protocols by identifying relevant multi-region electrical dynamics. Our approach models these dynamics as a superposition of latent networks, where the latent variables predict a relevant outcome. We use supervised autoencoders (SAEs) to improve predictive performance in this context, describe the conditions where SAEs improve predictions, and provide modelling constraints to ensure biological relevance. We experimentally validate our approach by finding a network associated with stress that aligns with a previous stimulation protocol and characterizing a genotype associated with bipolar disorder.

Keywords: dimensionality reduction; factor analysis; joint model; neuroscience; supervised autoencoders.

Abstract

Grants and funding