S1 Tools for the future of synaptic neuroscience: superresolution imaging meets artificial intelligence
Tobias Moser (Göttingen), Silvio Rizzoli (Göttingen) and Claudia Steinem (Göttingen)
Live Discussion: Monday, March 22, 2021, 17:00 - 18:00h
Synapses are the central processors of information in the brain. They are small and protein-dense compartments, whose molecular organization reflects their functional state. They have therefore been analyzed by imaging tools for decades.
Synaptic imaging has become increasingly complex, and includes many optical approaches, from wide-field imaging to nanoscale microscopy. Recent efforts focus on the large-scale imaging of multiple targets over large areas, and across long time intervals. Many challenges need to be met. Optical tools need to be developed and optimized to account for higher speed, multiplexing, three-dimensionality and for low phototoxicity. An additional challenge is the immense data load that is obtained. This surpasses the capacity of conventional expert-driven analysis, and therefore requires tools such as artificial intelligence.
Our symposium presents solutions for these challenges. We open with a technical view of the applications of optimized super-resolution imaging in small cellular compartments (Sauer). However, the emphasis on super-resolution optics is only useful when it is also coupled to substantial developments in the way the samples are analyzed. The analysis is especially a problem in synapses, where the molecules are densely packed, and therefore single-molecule information is especially difficult to obtain. New tools are approaching this problem (Cox), obtaining statistically-precise information from complex single-molecule maps. Building on these and related technologies, the dynamics of synaptic molecules are now analyzed, such as the AMPA glutamate receptors (AMPAR, presented by Choquet). To enable scientists to obtain such data with sufficient accuracy, pipelines have been developed for the high throughput acquisition of super-resolution images of synaptic proteins. These pipelines are now being generalized, through machine-learning approaches (Lavoie-Cardinal).
Finally, the field is also combining super-resolution imaging with electron microscopy and other approaches, to obtain information from the millisecond (Reshetniak) to the week (Lange) time scales.
Supported by SFB 1286 “Quantitative Synaptology”.