Recursion pharmaceuticals on GCP – A Success Story!

Cloud Computing

About the company

Recursion Pharmaceuticals uses technology at each step in drug discovery. The company headquartered in the Salt Lake City was founded in 2013, combines artificial intelligence with bioinformatics and experimental biology in a hybrid lab-to-cloud platform to identify treatments for any disease that Can be modeled at the cellular level. The main aim of the company is to build a map of human cellular
biology to change the pace and scale at which the new treatments can benefit the patients.

Recursion Logo

The Background

Despite the advancements in medical technology and scientific research, the drug discovery process continues to be at a slower pace and has become more expensive over the years. While the pharmaceutical companies are now spending more money on R&D each year, the results have not been that great. There is no increase in the number of new medicines that are FDA approved. Recursion is
looking to address this declining productivity by merging latest in machine learning with rich biological datasets to reinvent the development and drug discovery process.

This was when Recursion decided to choose Google Cloud Platform as their primary public cloud provider
as they continue to build their drug discovery platform. This platform combines automated biology, chemistry, and cloud computing to disclose new therapeutic candidates, possibly reducing the time to develop and discover a modern medicine by a factor of 10.

Microscope

The Pre-requisites

To accomplish this mission, the company developed a massive data pipeline which included inference engines, image processing, and in-depth learning modules that support bursts of computation power and integrated it with neural networks. With the help of this, Recursion has been able to create numerous disease models and created a shortlist of drug candidates across many diseases.

Beginning with wet biology

The plates of glass-bottom wells that consisted of thousands of diseased and healthy human cells. The biologists run experiments on these cells, apply stains that help them quantify and characterize the features of cellular samples: the thickness of the membrane, the shape of their mitochondria, their roundness and other characteristics. Automated microscopes at capture high-resolution photos
various light wavelengths.

Accelerating the drug discovery with GCP’s solution

Recursion was using the model that leveraged Confluent Kafka, and hence, they selected GCP since it has the most excellent Kubernetes. The Google Kubernetes Engine and the Confluent Kafka running on GCP analyzes and extracts cellular features from images. After that, the data is administered by deep neural networks to identify patterns containing that human may not be able to locate.

To match diseased and healthy cell signatures with those of cells after and before a variety of drug treatments, there are trained neural nets, therefore, promising a new potential therapeutics.

Moving from GPUs to TPUs

Recursion uses on-premise GPUs to train its deep learning models. They use Google Cloud Platform CPUs to make inference on new images in the pipeline. The company is presently evaluating cloud-based alternatives comprised of using Cloud TPU technology to automate and accelerate image processing. As the company was already using TensorFlow to train its neural networks, Cloud TPUs becomes a natural fit.

Cloud - Kubernetes Process

Stronger cloud integration

Recursion is also exploring using the Google Kubernetes Engine on-premise which is the foundation of the cloud services platform to manage all Kubernetes clusters from easy-to-use and single console.

 

Cloud partnering and open-source

The Google team stood out from other cloud providers in terms of the responsiveness and customized approach. The company founded Google’s Kubernetes solutions mature and a perfect fit for their environment. Recursion also found Google’s web console and the CLI more intuitive and ergonomic, which were apt for deep learning. The faster storage and Cloud TPU solution provided by Google helped them inch closer to their vision. Besides that, the open-source community of Google was the icing on the cake since the company from the start took an open-source approach for building its solution.

The Results

1.  With TPUs they could accelerate cellular microscopy image processing.

2. TensorFlow helped the company in cutting down deep learning model training from hours to
minutes.

3. Company’s local and cloud operations could be integrated on-premises with Google Kubernetes
Engine.

Google Cloud Platform resonated with Recursion’s vision to make it a success

Google is delighted to work in partnership with Recursion in their mission to inexpensively yet increasingly discover new medicines for scores of diseases that are both common and rare. You can learn more about this success story here.

Google also offers solutions for life sciences organizations, and you can also learn about Google Startup Programs Here

 

Now Become A Networking Expert by Learning  from  Operating System Interoperability Bundle

Leave a Reply

Your email address will not be published. Required fields are marked *