Title

AWS re:Invent 2022 - Faster drug discovery design with WEKA on AWS (PRT107)

Summary

The session focused on accelerating drug discovery using Weka on AWS.
Pruitt Chamness, director of cloud sales, and Baris Kiler, HPC practice manager at Clovertex, presented the talk.
Clovertex builds scalable, reliable, and performant research clouds for the pharmaceutical industry.
Cryo-EM (cryo-electron microscopy) is central to drug discovery, allowing the 3D structure of proteins to be determined for drug targeting.
The architecture used for the project includes AWS services and on-premises data generation, with Weka facilitating data transfer to the cloud.
Traditional on-premises workflows are being replaced by cloud-based solutions, offering flexibility and scalability.
Weka optimizes performance across different stages of the drug discovery pipeline without the need for tuning or data copying.
Benchmarks were conducted using CryoSpark software and the Empire 10288 dataset, simulating a real-world scenario with multiple users.
The results showed that Weka maximized CryoSpark software performance and GPU utilization.
Cost-performance analysis recommended G5 4X Large instances for the lowest cost and G5 24X Large for a balance of cost and speed.
The project demonstrated faster data transfer, independent scaling of capacity and throughput, quicker data availability, and simplified management.
The outcome for drug discovery companies is faster insights, quicker time to market, and scientists focusing on science rather than infrastructure.

Cloud-based workflows for drug discovery are becoming more prevalent, offering significant advantages over traditional on-premises solutions.
Weka's ability to handle different I/O profiles without tuning simplifies the data pipeline and accelerates drug discovery processes.
The use of AWS services and Weka allows for dynamic scaling, which is crucial for handling variable computational demands in drug research.
The benchmarking approach taken by Clovertex aimed to reflect real-world usage by simulating multiple concurrent users, providing more relevant performance data.
The cost-performance analysis indicates that selecting the right AWS instance types can significantly impact both the cost and speed of drug discovery workflows.
The transition to cloud-based solutions like Weka on AWS can lead to better resource utilization, such as more efficient use of expensive GPU resources.
The overall impact of using Weka on AWS for drug discovery is not only technical but also has significant business value, leading to faster patient outcomes and more efficient use of research budgets.