Image-centric compression of protein structures improves space savings

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Rapid data generation necessitates efficient compression algorithms for bioinformatics.
• Current methods often compress protein atomic coordinates sequentially, limiting compressibility.
• Novel approaches like Foldcomp focus on local angles, showing improved compression.

Purpose of the Study:

• To explore an image-based compression paradigm for 3D protein structures.
• To develop and evaluate a prototype compressor, PIC, utilizing global angles.
• To assess the effectiveness of mapping 3D atomic coordinates to 2D images for compression.

Main Methods:

• Implemented a prototype compressor named PIC.
• Mapped 3D atomic coordinates from PDB and mmCIF files to 2D greyscale images.
• Leveraged the PNG image compressor to reduce image file size.

Main Results:

• PIC achieved superior compression ratios over gzip for proteins larger than 20,000 atoms.
• Compression savings reached up to 37.4% compared to gzip.
• PIC's compression efficiency increased with protein size.

Conclusions:

• Image-centric compression, as demonstrated by PIC, offers a promising approach for 3D protein structure data.
• This method has the potential for developing specialized protein compression software.
• Further research is needed to optimize and implement this technique practically.