IsoNet

IsoNet reconstructs missing-wedge information in cryogenic electron tomography (cryoET) tomograms using deep learning to improve resolution isotropy and signal-to-noise ratio.

Key Features:

• Deep learning iterative reconstruction: Uses advanced deep learning algorithms to iteratively reconstruct missing information in cryoET tomograms.
• Resolution isotropy improvement: Enhances resolution isotropy to reduce anisotropic resolution artifacts caused by the missing-wedge.
• Signal-to-noise ratio enhancement: Improves signal-to-noise ratio in tomograms to increase structural interpretability.
• No sub-tomogram averaging required: Performs reconstruction without relying on sub-tomogram averaging workflows.
• Reduces resolution anisotropy: Directly addresses and decreases resolution anisotropy in reconstructed volumes.
• Supports downstream sub-tomogram averaging: Aids identification of complexes with differing orientations to support near-atomic resolution sub-tomogram averaging.
• Validated on diverse cryoET data: Demonstrated effectiveness across multiple types of cryoET datasets, including high-resolution tomograms.

Scientific Applications:

• Immature HIV lattice analysis: Resolves lattice defects within immature HIV particles to improve structural interpretation.
• Paraflagellar rod architecture: Elucidates the architecture of the paraflagellar rod in eukaryotic flagella from cryoET data.
• Clathrin cage identification in neurons: Identifies heptagon-containing clathrin cages within neuronal synapses of cultured cells.
• High-resolution tomogram enhancement: Enhances interpretability of high-resolution cryoET tomograms to facilitate functional analysis of cellular components.

Methodology:

Iterative deep learning–based reconstruction of missing-wedge information in cryoET tomograms, performed without sub-tomogram averaging.