Bayesian inference of molecular kinetic parameters from astrocyte calcium imaging data

Abstract

Model-based Bayesian inference from high-content data obtained on live specimens is a burgeoning field with demonstrated applications to neuroscience. In parallel, computer vision methods for extracting the calcium signaling information from imaging data have advanced in application to neuronal physiology. Here, we are describing in detail a method we have recently developed to study calcium dynamics in astrocytes, which combines computer vision with model-based Bayesian learning to deduce the most likely molecular kinetic parameters underlying the observed calcium activity. As reported in the companion experimental study, this method allowed us to identify the key molecular changes downstream of a multi-gene deletion modeling the human 22q11.2 deletion syndrome, the most common human microdeletion and the genetic factor with the highest penetrance for schizophrenia. • Methodological details are laid out, from our imaging approach to our adaptation of the VBA-CaBBI algorithm previously developed primarily for brain functional imaging data. • The analytical pipeline is suited for further applications to glial cells and adaptable to other cell types exhibiting complexcalcium dynamics.

Keywords

• Schizophrenia
• 22q11.2 deletion syndrome
• Glia
• Model-based inference
• Li-Rinzel model
• Calcium dynamics