Long-term neural and physiological phenotyping of a single human

Russell A. Poldrack, Timothy O. Laumann, Oluwasanmi Koyejo, Brenda Gregory, Ashleigh M. Hover, Mei-Yen Chen, Krzysztof J. Gorgolewski, Jeffrey J. Luci, Sung Jun Joo, Ryan L. Boyd, Scott Hunicke‐Smith, Zack B. Simpson, Thomas Caven, Vanessa Sochat, James M. Shine, Evan M. Gordon, Abraham Z. Snyder, Babatunde Adeyemo, Steven E. Petersen, David C. Glahn, Dean McKay, Joanne E. Curran, Harald H.h. Göring, Melanie A. Carless, John Blangero, Robert F. Dougherty, Alexander Leemans, Daniel A. Handwerker, Laurie Frick, Edward M. Marcotte, Jeanette A. MumfordNature Communications 2015 · DOI

TL;DR

This paper introduces the "MyConnectome" project, an unprecedentedly intensive, 18-month phenome-wide study of a single individual. It reveals dynamic changes in brain connectivity, psychological states, gene expression, and metabolism over daily to monthly timescales, demonstrating that human physiology is far more dynamic than previously assumed. This work provides a valuable, open-access dataset and reproducible workflow for studying individual variability and its implications for brain disorders.

What Problem Does This Paper Address?

While we understand brain function dynamics over seconds (learning) and years (development/aging), the intermediate timescale of days to months in healthy individuals remains largely unexplored. This is a critical gap, especially considering that major psychiatric disorders often manifest with significant fluctuations in psychological and neural function over these periods. The challenge lies in collecting dense, longitudinal data from healthy volunteers, who are unlikely to sustain the required commitment. This study tackles this by performing an exceptionally detailed, long-term phenotyping of a single, motivated individual (one of the authors) to create a proof-of-concept for such intensive studies.

Methods

The MyConnectome project employed a comprehensive suite of methods over 18 months:

  • Neuroimaging:
    • Resting-state fMRI (rsfMRI): The most frequently collected data (84 usable sessions), used to map functional brain connectivity.
  • Structural MRI (T1/T2-weighted): To map brain anatomy.
  • Diffusion-weighted MRI (DTI): To map structural connectivity.
  • Task-based fMRI: For specific cognitive tasks (n-back, stop-signal, object localization, spatial working memory, language localization, retinotopic mapping).
  • Physiological and Biological Measurements:
    • Psychological assessments: Daily surveys capturing mood, fatigue, and other self-reported states.
  • Gene expression: RNA sequencing of peripheral blood mononuclear cells (PBMCs) to assess genome-wide gene expression.
  • Metabolomics: Gas chromatography time-of-flight mass spectrometry of blood serum to quantify metabolite concentrations.
  • Data Analysis:
    • Reproducible workflow: A Python package and virtual machine were developed to allow others to download data and reproduce analyses.
  • Network analysis: Graph-theoretic approaches were used to analyze functional and structural brain connectivity, including modularity and global efficiency.
  • Time-series analysis: Linear regression with automated model order selection was used to identify associations between different data types over time.
  • Weighted Gene Coexpression Network Analysis (WGCNA): Used to identify modules of coexpressed genes.
  • Phenome-wide network analysis: A network was constructed where nodes represented different variables (brain connectivity, gene expression, metabolites, behaviors) and edges represented significant associations between them.

Datasets

The primary dataset is the MyConnectome dataset, collected from a single healthy human male over 18 months. This dataset includes:

  • Functional and structural MRI scans.
  • Psychological self-report data.
  • RNA sequencing data from PBMCs.
  • Metabolomics data from blood serum.

The paper also references the GOBS (Genetics of Brain Structure and Function) study dataset for validating gene expression module heritability.

Key Insights and Ideas

  • Human physiology is highly dynamic: Contrary to the assumption of relative stability, brain connectivity, gene expression, and metabolic profiles exhibit significant fluctuations on timescales of days to months.
  • Interconnectedness of systems: Brain function is not isolated but intricately linked with psychological states, gene expression, and metabolic processes. Changes in one system often correlate with changes in others.
  • Food and caffeine impact brain networks: Even short-term physiological states like being fed/caffeinated versus fasted/uncaffeinated can alter the structure and connectivity of large-scale brain networks.
  • Individual variability is key: The study highlights the importance of understanding within-individual variability, which may be as crucial as between-individual differences for understanding health and disease.
  • Open science is powerful: The commitment to open data and reproducible workflows is central to the project's value, enabling broader scientific inquiry.

Key Results

  • Dynamic Brain Connectivity: rsfMRI data revealed dynamic changes in functional connectivity, particularly in visual, somatomotor, and attention networks, showing linear or polynomial trends over the study period.
  • Phenome-Wide Associations: A phenome-wide analysis identified numerous significant associations between brain connectivity, gene expression modules, metabolites, and behavioral measures. For example, alcohol intake was strongly associated with the metabolite butane-2,3-diol, and specific brain network connectivities were linked to fatigue.
  • Gene Expression and Brain Connectivity: While within-network connectivity showed fewer associations, between-network connectivity demonstrated a notable number of strong relationships with gene expression modules.
  • Metabolite-Food Intake Link: Metabolomic variables showed strong relationships with food intake, with specific examples like bacon intake correlating with ketone body concentrations.
  • Structural vs. Functional Connectivity: Regularized partial correlation measures of functional connectivity showed a better match to structural connectivity (DTI) than full correlation, suggesting that partial correlation better captures direct structural links.

Limitations and Future Work

  • Single Subject: The most significant limitation is that the study is based on a single individual. While this provides an unprecedented depth of data, it limits generalizability. The authors acknowledge this and state that the findings are exploratory and generate hypotheses for future studies with larger, unbiased samples.
  • Self-Experimentation Bias: The authors acknowledge the potential for subject expectations to influence behavioral and neural measures, although they took steps to mitigate this. They argue that it's harder to explain how expectations would drive the observed gene expression and metabolomic associations.
  • Reproducibility of Analyses: While the workflow is provided, the computational demands for some preprocessing steps (e.g., imaging, genomics) might still require high-performance computing.
  • Future Work: The authors emphasize that this study is a proof-of-concept and a testbed for future research. They suggest that larger longitudinal studies are crucial to test the generated hypotheses, particularly for understanding psychiatric disorders characterized by significant temporal variability. The open data resource is intended to facilitate these future investigations.

Summary updated

Abstract

Psychiatric disorders are characterized by major fluctuations in psychological function over the course of weeks and months, but the dynamic characteristics of brain function over this timescale in healthy individuals are unknown. Here, as a proof of concept to address this question, we present the MyConnectome project. An intensive phenome-wide assessment of a single human was performed over a period of 18 months, including functional and structural brain connectivity using magnetic resonance imaging, psychological function and physical health, gene expression and metabolomics. A reproducible analysis workflow is provided, along with open access to the data and an online browser for results. We demonstrate dynamic changes in brain connectivity over the timescales of days to months, and relations between brain connectivity, gene expression and metabolites. This resource can serve as a testbed to study the joint dynamics of human brain and metabolic function over time, an approach that is critical for the development of precision medicine strategies for brain disorders.
BibTeX
@inproceedings{poldrack2015long,
  author = {Russell A. Poldrack and Timothy O. Laumann and Oluwasanmi Koyejo and Brenda Gregory and Ashleigh M. Hover and Mei-Yen Chen and Krzysztof J. Gorgolewski and Jeffrey J. Luci and Sung Jun Joo and Ryan L. Boyd and Scott Hunicke‐Smith and Zack B. Simpson and Thomas Caven and Vanessa Sochat and James M. Shine and Evan M. Gordon and Abraham Z. Snyder and Babatunde Adeyemo and Steven E. Petersen and David C. Glahn and Dean McKay and Joanne E. Curran and Harald H.H. Göring and Melanie A. Carless and John Blangero and Robert F. Dougherty and Alexander Leemans and Daniel A. Handwerker and Laurie Frick and Edward M. Marcotte and Jeanette A. Mumford},
  title = {{Long-term neural and physiological phenotyping of a single human}},
  booktitle = {Nature Communications},
  year = {2015},
  doi = {10.1038/ncomms9885},
}
Figure