Hong Kong Baptist University (HKBU) Research Cluster on Data Analytics and Artificial Intelligence in X

Robustness and sloppiness in human brain networks: reliability and variability across individuals, aging and genetic modulations
Principal Investigatgor: Prof. Changsong ZHOU ( Department of Computer Science )
Co-Investigatgor: Dr. Longjiang Zhang ( Jinling Hospital, Medical School, Nanjing University )
Co- Investigatgor: Prof. Andrea Hildebrandt ( Department of Psychology, University of Oldenburg, Germany )

Human cognitive performance is based on dynamical patterns organized through complex interconnected brain networks which are developed through long time evolution. Recent research has paid much attention to structure-function relationship within brain networks by comparing different groups of participants – mostly persons in pathological conditions compared with typically functioning individuals. However, typically functioning individuals also substantially differ in cognitive behavior, but such individual differences as rooted in the brain’s complex structural and functional networks are only poorly understood to date. Observable individual differences may deeply relate to the evolution principle of a trade-off between evolutionary robustness by developing a reliable structure and function and evolvability through strong enough phenotype variability. Several crucial questions have not yet been well addressed. How reliable and variable are the brain’s structural and functional networks across individuals? Is the reliability/variability uniform or heterogeneous within the network? How is the measured variability related to behavioral differences between individuals? Why some variations in the brain networks lead to typical behavior and others to pathological conditions? How are typical development and aging, and brain disorder and disease related with the brain network reliability and variability? Address these questions is of fundamental importance to understanding brain normal and abnormal functions and developing engineering and medical modulations. Especially, variation in brain disorders and diseases may happen at the extreme ends of the variability spectra, and early detection becomes highly challenging when masked by typical individual differences. It is highly desirable to develop a systematic theoretical/computational framework for interpersonal variability and validate in large samples.

This collaborative project aims to address the above fundamental questions by combining large-scale multimodal brain imaging datasets (~2000 subjects) with computational model in a novel framework of sloppiness in systems biology. Sloppiness refers to an anisotropic parameter space, wherein the system state is highly sensitive to variation along some ‘stiff’ directions (combinations of parameters) and less sensitive in many ‘sloppy’ directions. The possible stiffness and sloppiness in the brain network will be studied using pairwise maximal entropy (Ising) model fitted to the brain resting state spatiotemporal activity patterns measured from functional MRI; The stiffness and sloppiness will be directly compared with actual interindividual reliability and variability of the brain networks measured from MRI across large samples of individuals and used to predict behavior performance. This framework may provide novel approaches to identify fingerprints of individuals in cognitive performance as the combinations of different stiff or sloppy structural links. We will also study how the brain networks change with aging and how these changes are modulated by risk gene APOE ε4 of Alzheimer’s Diseases (AD). We hypothesize that normal aging and atypical neurodegeneration may target network subsets with different stiffness levels. Potential findings in this project may shed fundamentally new lights into the organization and universal working principle of the brain networks underlying stable function and diverse individual differences and contribute to the question of how these principles transfer to the extreme ends of brain disorders. Thus, the project will build foundation for future systematic studies of various types of brain diseases to potentially identify correspondingly the most crucial brain components for electrical or magnetic modulations for intervention and prevention in engineering and clinical applications.


  • Process multimodal brain MRI data to obtain structural measures and dynamical neural activity patterns for large sample of participants.
  • Fit Ising model to fMRI data and obtain Fisher information matrix (FIM) for each participant. Perform eigenmode analysis of FIM to identify stiff and sloppy directions and analyze anisotropy of the parameter space.
  • Characterize the spectra of interindividual reliability and variability in brain networks at different stiffness level from experimentally measured parameters in Objective (1).
  • Investigate how the network changes with aging and in function of genetic risk to developing AD are related to stiffness and sloppiness.

Grant Support:

This project is supported by the Research Grants Council (RGC), Hong Kong SAR, China (Project 12301019).

For further information on this research topic, please contact Prof. Changsong ZHOU.