Beyond fingerprinting: Choosing predictive connectomes over reliable connectomes

Area of Science:

• Neuroscience
• Cognitive Science
• Psychology

Background:

• Recent research focuses on functional brain connectivity variability.
• Concerns exist regarding the reliability and reproducibility of functional connectivity findings.
• There is a need for robust methods to assess functional connectivity.

Purpose of the Study:

• To advocate for using behavior prediction as the primary benchmark for functional connectivity research.
• To address concerns about reliability in the context of inter- and intra-subject variability.
• To guide the neuroimaging community toward more applicable and reproducible research.

Main Methods:

• Theoretical arguments and empirical evidence are presented.
• The opinion piece discusses trait-level differences and state-related changes.
• Focus is placed on predictive modeling for real-world applications.

Main Results:

• Optimizing reliability for its own sake is less valuable than predicting behavior.
• Behavior prediction serves as a more meaningful benchmark for functional connectivity.
• Characterizing inter- and intra-subject variability is crucial for predictive models.

Conclusions:

• Behavior prediction should be the benchmark for functional connectivity research, not just reliability.
• This perspective aids in understanding individual differences and state changes in brain function.
• The approach supports building predictive models for practical applications in neuroscience and psychology.