How PTP Enabled 12x Faster AlphaFold Workflows and 86% Cost Savings* with AWS HealthOmics

Illustration of Goat working on servers leading data to the cloud and to a proved treatment

A biotechnology firm in stealth mode, dedicated to advancing vaccines for global health, leveraged AlphaFold to predict and understand protein structures using machine learning. AlphaFold, powered by convolutional neural networks (CNN), offers groundbreaking insights but requires extensive computational resources. Before engaging PTP, the company faced severe limitations with their on-premises infrastructure, hindering their ability to scale AlphaFold effectively.

%

*The HealthOmics deployment resulted in an 86% cost reduction compared to the initial proof of concept deployment using AWS.

Freed from static on-prem constraints, the company’s workflows accelerated by 12 times.

The Challenge

The company’s existing infrastructure presented several roadblocks:

Limited GPU Access
AlphaFold was deployed on individual laptops or a single server with restricted GPU capabilities. Their laptops lacked GPU provisioning, and the server’s GPU capacity was physically constrained.

Insufficient Storage
The server lacked adequate storage to process lab data simultaneously, creating significant bottlenecks.

On-Prem Environment
The company’s reliance on on-premises infrastructure severely limited scalability, and they had no prior experience using AWS.

Unpredictable Workload Scaling
The complexity of AlphaFold’s computational requirements made predicting resource needs difficult, rendering additional on-prem purchases financially unviable.

Lack of Cloud Expertise
The team lacked the technical expertise to implement and scale AlphaFold workflows in the cloud.

To progress toward their research targets, the company needed a scalable solution to run AlphaFold efficiently, reduce costs, and accelerate processing times.

The Solution 

PTP introduced a cloud-scale computing platform to address the company’s challenges and accelerate their research goals:

Initial Proof of Concept
PTP migrated the company’s data to Amazon S3 and deployed AlphaFold on traditional EC2 compute resources to demonstrate cloud viability.

This proof of concept validated that AWS could match or exceed their on-prem performance.

Optimized Deployment
PTP evaluated two deployment options for AlphaFold: AWS Batch for a bespoke deployment and AWS HealthOmics.

The company selected HealthOmics due to its advanced security features, robust storage capabilities, and native integration with NextFlow for managing workflows.

Custom HealthOmics Deployment
HealthOmics supported the deployment of AlphaFold’s “Ready to Run” version but required customization for the company’s complex project, which involved processing more than 12,000 residues (compared to the standard 1,200).

PTP adjusted existing scripting and deployment code within HealthOmics, enabling faster development and scalability.

Scalable Infrastructure
PTP’s deployment provided the company with access to multiple servers and GPUs, drastically reducing computational times from days to hours.

Infrastructure Diagram

AWS reference architecture diagram illustrating the workflow for AlphaFold optimization using AWS Batch, AWS CloudFormation, Amazon FSx for Lustre, and AWS Code services. The diagram showcases job inputs and results moving through a Virtual Private Cloud (VPC) with general and GPU compute instances, integrating public datasets and models.

Figure 1: Reference Architecture for Proof-of-Concept Deployment for The company

AWS HealthOmics workflow diagram showing how users interact with Amazon S3 for job inputs and results, processing genomic data through AWS HealthOmics for optimized storage and analysis.

Figure 2: Reference Architecture for Custom HealthOmics Deployment 

The Outcome


PTP’s engagement delivered transformative results:

12x Faster Workflows
Freed from static on-prem constraints, the company’s workflows accelerated by 12 times.

86% Cost Savings
The HealthOmics deployment resulted in an 86% cost reduction compared to the initial proof of concept deployment using AWS.

Massive Cost Avoidance
Transitioning from on-prem infrastructure eliminated the need for significant equipment purchases, maintenance costs, and man-hours.

Improved Scalability
The custom HealthOmics deployment allowed the company to scale AlphaFold workflows efficiently and process highly complex data sets.

Graphs Isometric Contained Icon

Ready to scale your computational workflows?

By leveraging PTP’s cloud expertise and AWS HealthOmics, the company significantly reduced costs, accelerated research timelines, and optimized functionality. Contact PTP today to learn how we can help optimize your cloud environment for research and innovation.

 

Let us help you unlock your potential.

Contact PTP today to learn how we can deliver cost-efficient, scalable, and compliant cloud solutions for your business.

Homepage Contact Us