Here we report SUPPORT (statistically unbiased prediction utilizing spatiotemporal information in imaging data), a self-supervised learning method for removing Poisson-Gaussian noise in voltage imaging data. SUPPORT is based on the insight that a pixel value in voltage imaging data is highly dependent on its spatiotemporal neighboring pixels, even when its temporally adjacent frames alone do not provide useful information for statistical prediction. Such dependency is captured and used by a convolutional neural network with a spatiotemporal blind spot to accurately denoise voltage imaging data in which the existence of the action potential in a time frame cannot be inferred by the information in other frames. Through simulations and experiments, we show that SUPPORT enables precise denoising of voltage imaging data and other types of microscopy image while preserving the underlying dynamics within the scene.

Biological systems consist of hierarchical protein structures, each of which has unique 3D geometries optimized for specific functions. In the past decades, the growth of inorganic materials on specific proteins has attracted considerable attention. However, the use of specific proteins as templates has only been demonstrated in relatively simple organisms, such as viruses, limiting the range of structures that can be used as scaffolds. This study proposes a method for synthesizing metallic structures that resemble the 3D assemblies of specific proteins in mammalian cells and animal tissues. Using 1.4 nm nanogold-conjugated antibodies, specific proteins within cells and ex vivo tissues are labeled, and then the nanogold acts as nucleation sites for growth of metal particles. As proof of concept, various metal particles are grown using microtubules in cells as templates. The metal-containing cells are applied as catalysts and show catalytic stability in liquid-phase reactions due to the rigid support provided by the microtubules. Finally, this method is used to produce metal structures that replicate the specific protein assemblies of neurons in the mouse brain or the extracellular matrices in the mouse kidney and heart. This new biotemplating approach can facilitate the conversion of specific protein structures into metallic forms in ex vivo multicellular organisms.