Brain controllability: Not a slam dunk yet.

Area of Science:

• Neuroscience
• Network Science
• Computational Biology

Background:

• The controllability of brain networks is crucial for understanding brain function.
• Previous studies, such as Gu et al. [2], proposed methods to assess node contributions to controllability.
• Concerns have been raised regarding the robustness and interpretability of these methods.

Purpose of the Study:

• To provide quantitative evidence highlighting warnings and caveats in the methodology for defining and measuring brain network controllability.
• To address the limitations of assessing one-node controllability using numerical analysis.
• To investigate the reliability of the framework proposed by Gu et al. [2].

Main Methods:

• Replication and extension of the methodology presented in Gu et al. [2].
• Application of controllability analysis to multiple human brain network datasets.
• Numerical investigation of one-node controllability using the same framework as Gu et al. [2].

Main Results:

• The minimum eigenvalue of the controllability Gramian (λmin(WK)) was found to be statistically compatible with zero.
• The controllability Gramian could not be inverted, indicating numerical instability.
• These findings demonstrate that one-node controllability of the brain cannot be reliably inferred numerically.

Conclusions:

• The methodology for assessing one-node brain network controllability, as presented by Gu et al. [2], has significant limitations.
• Numerical analyses of controllability in brain networks often lead to ill-conditioned problems, yielding unreliable results.
• Both the current study and Pasqualetti et al. [3] agree that numerical inference of one-node controllability is not feasible.