GoToCloud: Development and future of the cloud-based platform for cryo-EM structure-based drug DesignDevelopment and...

Area of Science:

• Structural Biology
• Computational Chemistry
• Biophysics

Background:

• The
• Resolution Revolution
• in cryogenic electron microscopy (Cryo-EM) has significantly advanced structure-based drug design (SBDD).
• Exponential data growth from Cryo-EM creates computational bottlenecks, exceeding the capabilities of traditional on-premise high-performance computing (HPC).
• Existing HPC solutions lack the scalability and economic flexibility required for modern SBDD workflows.

Purpose of the Study:

• To introduce "GoToCloud," a cloud-based platform designed to address computational and data transfer challenges in Cryo-EM.
• To enable researchers to deploy secure, scalable, and cost-effective virtual HPC clusters for SBDD.
• To facilitate rapid and efficient global data mobility for distributed Cryo-EM analysis.

Main Methods:

• Leveraged Amazon Web Services (AWS) ParallelCluster to create secure virtual clusters within a Virtual Private Cloud (VPC).
• Implemented a "Shared EFS" architecture with automated scripts for simplified cluster deployment and software pre-installation (e.g., RELION).
• Integrated Zettar zx for high-speed, intercontinental data transfer, utilizing burst-buffer architecture to overcome on-premise bottlenecks.

Main Results:

• The GoToCloud platform demonstrated effective and scalable Cryo-EM data analysis, achieving 1.83Å resolution.
• Economic analysis identified optimal GPU instance types (NVIDIA T4 vs. A10G) for balancing cost and performance, reaching the "2.0Å wall."
• Achieved high-throughput data transfer rates of ~4.6 Gbps over intercontinental distances, significantly faster than physical media transport.

Conclusions:

• GoToCloud democratizes access to HPC resources for Cryo-EM researchers, eliminating the need for specialized IT maintenance.
• The platform establishes a robust foundation for next-generation SBDD, enabling automated high-throughput screening and validation of AI-predicted models.
• Enhanced data mobility and computational scalability accelerate the drug discovery pipeline.