From Years to Days: Transforming Drug Discovery with NVIDIA DGX and OCI AI Innovations

Outcomes:

Increased Productivity:

Developer productivity experienced a substantial 50% boost, driven by the enhanced efficiency and capabilities provided by DGX Cloud.

Time Savings:

The setup time for multi-node training, which previously took 7-10 months, has now been significantly reduced, greatly improving the overall process and efficiency.

Efficient Pipeline Assembly:

What used to take 4-6 weeks to assemble can now be completed in just a few clicks, greatly accelerating research workflows and enhancing overall efficiency.

Introduction:

In the quest to revolutionize drug discovery, leveraging advanced technologies is crucial. This case study explores how CSTX and its partner TATA helped a leading pharmaceutical enterprise enhance its drug discovery process through an innovative AI platform using NVIDIA DGX Cloud on Oracle Cloud Infrastructure (OCI).

Client Overview:

Client: Pharmeceutical Enterprise
Industry: Healthcare and pharmaceutical
Scope: Accelerating drug discovery processes and improving drug development efficiency.

Challenges:

The client faced several challenges in expanding their data center operations globally:

Drug discovery is a traditionally lengthy and costly process, often taking 10–15 years and costing over $1–2 billion per drug.
The high failure rate of 90% in drug trials exacerbates these challenges.
The client needed to address these issues by enhancing preclinical models, improving target validation, and refining decision-making strategies to reduce trial failures and streamline the drug development journey.

Objectives:

The primary objective was to expedite the drug discovery process using advanced AI technologies. CSTX and TATA aimed to deploy NVIDIA DGX Cloud on OCI to:

Accelerate drug discovery through generative AI and large language models (LLMs).
Utilize NVIDIA BioNeMo™ and other NVIDIA technologies to develop and train custom models for chemistry and protein language.
Seamlessly deploy and scale AI-driven solutions with NVIDIA NIM inference microservices.

Solutions:

The solution involved the deployment of an end-to-end NVIDIA DGX HPC infrastructure, including:

NVIDIA DGX Cloud for AI computation.
NVIDIA BioNeMo for generative AI and LLM applications.
Mellanox switch infrastructure and Infiniband networking for high-performance connectivity.
DDN Storage for data management.

Approach: CSTX and TATA implemented a comprehensive infrastructure at the client’s data center, setting up the necessary hardware and software components. They focused on leveraging the NVIDIA DGX Cloud to enhance AI capabilities, accelerate training times, and optimize drug discovery processes through advanced AI models.

Project Execution:

Infrastructure Setup: CSTX delivered a robust Nvidia DGX HPC solution, including cabling, compute resources, switching, and connectivity.
AI Model Training: Using NVIDIA BioNeMo’s LLM and generative AI models, the training time for AI-generated compounds was reduced from 4 weeks to 8 days.
Enhanced Insights: Multimodal datasets were employed to rank and refine molecules iteratively using NVIDIA MolMIM.

Conclusion:

The integration of NVIDIA DGX Cloud and associated AI technologies has substantially transformed the client’s drug discovery process. By reducing training times, enhancing data insights, and improving overall productivity, the solution has paved the way for more efficient drug development and a reduction in costly trial failures. This case study highlights the significant impact of advanced AI infrastructure in addressing long-standing challenges in pharmaceutical research.