GPU-enabled FREALIGN: accelerating single particle 3D reconstruction and refinement in Fourier space on graphics...

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Single particle electron cryo-microscopy (cryoEM) is crucial for determining high-resolution 3D structures of biological macromolecules.
• Dataset size significantly impacts cryoEM 3D reconstruction resolution, with larger datasets generally yielding better results.
• Processing very large cryoEM datasets is computationally intensive and time-consuming.

Purpose of the Study:

• To accelerate the processing of large cryoEM datasets.
• To enhance the computational efficiency of the FREALIGN software.
• To enable higher resolution 3D reconstructions from extensive cryoEM data.

Main Methods:

• Modification of the FREALIGN program to incorporate Graphics Processing Unit (GPU) acceleration.
• Parallelization of computationally intensive subroutines within FREALIGN.
• Performance evaluation of the modified FREALIGN on multi-GPU computer systems.

Main Results:

• Significant speed increases, ranging from 10 to 240-fold, were achieved for critical FREALIGN tasks.
• The modified FREALIGN program effectively leverages multiple GPUs for accelerated computation.
• The parallelized subroutines demonstrate robust performance, facilitating the processing of larger datasets.

Conclusions:

• GPU acceleration dramatically improves the speed of cryoEM data processing using FREALIGN.
• The modified FREALIGN is a valuable tool for researchers working with large cryoEM datasets.
• This advancement supports the generation of higher resolution 3D reconstructions in structural biology.