A SIMULATION-BASED INFERENCE FRAMEWORK FOR DECIPHERING BIOPHYSICAL MECHANISMS OF PATHOLOGICAL BURSTING IN DUP15Q SYNDROME

This study introduces a computational framework to bridge the "translational gap" in Duplication 15q (Dup15q) syndrome, a genetic disorder linked to autism and epilepsy where the link between microscopic findings and clinical symptoms often remains qualitative. We implemented a high-fidelity biophysical bridge by coupling patient-derived human-induced pluripotent stem cell (hiPSC) neuronal networks with Simulation-Based Inference (SBI). Experimentally, we employed a the human-derived model by differentiating patient-specific hiPSCs into excitatory neurons via the forced overexpression of the transcription factor Neurogenin-2. These neurons were co-cultured in a 1:1 ratio with rat cortical astrocytes on 24-well Micro-Electrode Arrays (MEAs) to provide essential metabolic and structural support for network maturation. Our computational model, built using the Brian2 toolkit, utilizes Hodgkin-Huxley neurons, Tsodyks-Markram synapses, and an active astrocyte model to simulate realistic network dynamics. While control networks show stereotyped single-peak bursts, experimental results at DIV 35 reveal that Dup15q networks exhibit a pathological fragmented-burst phenotype. By employing Simulation-Based Inference (SBI), we map these fragmented phenotypes back to biophysical parameters, identifying altered kinetics in short-term synaptic depression or NMDA receptor activity as primary drivers of network dysfunction. We expect to demonstrate that this fragmentation is driven by altered kinetics in short-term synaptic depression or NMDA receptor activity. By unraveling complex genetic signatures into actionable biophysical targets, this methodology provides a quantifiable link between cellular mechanisms and macroscopic pathology, establishing a personalized platform for targeted drug screening and neuroprotective therapies.