Quantized Neural Representation for Lossy Cryo-EM Compression

Area of Science:

• Structural biology
• Biophysics
• Computational biology

Background:

• Cryo-electron microscopy (cryo-EM) generates large 3D density volumes crucial for structural biology.
• Increasing data scale challenges storage, transmission, and interactive visualization of cryo-EM data.
• Efficient data compression is needed to overcome these limitations.

Purpose of the Study:

• To propose a novel hybrid quantized implicit neural representation (INR) method for cryo-EM data compression.
• To enable efficient on-demand access to compressed cryo-EM volumes.
• To evaluate the compression efficiency and visual fidelity of the proposed method.

Main Methods:

• Implemented a hybrid quantized implicit neural representation (INR) approach.
• Benchmarked the INR method against traditional compression techniques and a classic INR compressor.
• Assessed compression efficiency, visual quality, and performance on key cryo-EM tasks (structure identification, secondary structure recognition, fine-chain inspection).

Main Results:

• The quantized INR method demonstrated superior storage efficiency compared to traditional methods.
• The approach maintained high task-relevant fidelity for cryo-EM data analysis.
• Performance benchmarks confirmed the effectiveness of the INR method for structural biology applications.

Conclusions:

• The hybrid quantized INR method offers a powerful solution for compressing large cryo-EM datasets.
• This technique facilitates efficient storage, transmission, and interactive exploration of 3D density volumes.
• An interactive tool and guidelines are provided to aid users in selecting optimal compression strategies.