GENERATIVE MODELS FOR LARGE-SCALE SIMULATIONS OF CONNECTOME DEVELOPMENT.

Area of Science:

• Neuroscience
• Graph Theory
• Computational Biology

Background:

• The brain's connectome, representing functional and anatomical connections, is mathematically modeled as a graph.
• Estimating connectome topology from functional data alone poses challenges in unraveling developmental and pathological processes.
• Biologically meaningful simulations require generative models that produce synthetic graphs with controllable, data-driven parameters.

Purpose of the Study:

• To present a novel, data-driven approach for parameter tuning of the Lancichinetti-Fortunato-Radicchi (LFR) generative model for connectome graphs.
• To apply this tuned model to generate large datasets of synthetic brain graphs representing different maturation stages.
• To gain insights into the topological organization changes during neural development.

Main Methods:

• Utilized a large dataset of resting-state functional magnetic resonance imaging (fMRI) connectomes (n=5566) from the Adolescent Brain Cognitive Development (ABCD) Study.
• Developed a data-driven method to tune the parameters of the LFR generative model based on real connectome data.
• Employed the tuned LFR model for simulations to generate synthetic brain graphs.

Main Results:

• Successfully tuned LFR model parameters using a large-scale adolescent brain connectome dataset.
• Generated large synthetic datasets of brain graphs simulating various stages of neural maturation.
• The simulations provide a basis for investigating developmental changes in brain network topology.

Conclusions:

• The proposed data-driven approach enhances the biological relevance of synthetic connectome graphs generated by models like LFR.
• This method facilitates the study of brain development and aging by providing realistic simulations.
• It offers a valuable tool for exploring the complex topological organization of the developing human brain.