Based in Berkeley and founded in 2022, WittGen Biotechnologies develops high-resolution, single-cell profiling software to help researchers and clinicians treat cancer patients. Analyzing solid tumor RNA sequencing data (focusing on heterogeneity analysis), the company’s machine learning algorithm curates sequenced patient single-cell data sets. The platform technology then predicts, characterizes, and categorizes tumor cell phenotypes and the genetic molecular subtypes of cancer—thus disentangling the intra-tumor heterogeneity of transcriptome profiles.
The fight to cure cancer has long been hindered by high heterogeneity, which strictly limits research to improve clinical outcomes and drug responsiveness. But the healthcare industry may be receiving a boost in taking on this challenge from Yun Rhie, Minwoo Jung, and Minjun Kim—the founders of WittGen Biotechnologies.
The company has developed a machine learning algorithm that profiles and categorizes tumor cell phenotypes by integrating multimodal single-cell data. With this capability, the WittGen platform will provide a method for cancer researchers to disentangle inter-tumor and intra-tumor heterogeneity of transcriptome profiles. The results of this research will enable clinicians to manage patient cases more effectively.
After launching in 2022 and operating in start-up mode, WittGen experts developed the front-end interface and the machine learning model for the platform running on servers in an on-premises data center. The platform analyzes tumor DNA samples and produces genomic data that helps with cancer research.
As company executives planned to go to market with the service, they needed to develop a machine learning back-end in the cloud to set up the platform as a software-as-a-service offering. The co-founders also wanted to leverage the cloud for the necessary computing power and to be able to scale services to handle the projected growth in user demand as the platform went live.
“The customers of our platform—cancer researchers, oncologists and pathologists—are not familiar with coding to run analyses, so we needed to set up a machine learning engine in the cloud for them to use,” explains Jung. “As a small start-up, we did not have back-end developers on staff, so we did not have the resource bandwidth to get it done on time to satisfy our investors—who were looking for us to go to market quickly. Building a cloud infrastructure was also something we were not familiar with.”
WittGen had already established a cloud environment with Amazon Web Services (AWS) and knew AWS was the best cloud platform to deliver the required application performance and security. As WittGen qualified for an AWS program covering the cost of developing the cloud environment, the next step was to find a partner with the necessary infrastructure and machine learning backend expertise.
AWS introduced WittGen to Avahi Technologies as a potential partner, and it didn’t take long for Jung, Rhie and Kim to realize Avahi could do the job. “Once we described our platform concept, they instantly understood what researchers and clinicians would need to use our service,” says Rhie. “Comprehending a complex technology offering so quickly does not always happen when working with third-party developers—but the Avahi team was in sync with our vision right away.”
Soon after the two companies met, Avahi produced a backend schema and AWS architecture to support the platform to confirm the designs were on target with what WittGen wanted. “It took them just a few days and matched our requirements exactly,” adds Jung. “We could see that partnering with Avahi would be just like collaborating with our own internal team—and that gave us confidence early in the relationship.”
The image below shows the AWS infrastructure Avahi designed and built to support the WittGen platform. It includes a portal researchers can log into to save DNA and genomic data pertaining to cancer patients on Amazon S3 (Simple Storage Service).
Each time a request comes in, a WittGen data officer first pre-processes the data on an Amazon EC2 (Elastic Cloud Compute) server to perform quality control and format the data for machine learning analysis. The AWS Batch service then uses the containerized machine learning model to run computing jobs in Amazon DynamoDB and saves the analysis output to S3.
Clinicians can then log into the portal to view and download the results as they consider treatments for their cancer patients:
The platform relies on Amazon CloudWatch to monitor the performance of all compute resources and AWS services. Avahi also applied security controls, such as Amazon Cognito and AWS IAM (Identity Access Management) to ensure access to the platform is limited to authorized users.
The value Avahi delivered by designing and deploying the machine learning platform is demonstrated on the WittGen website. The company displays the schema image to prove the strength of the platform to the healthcare industry.
During the project, Yung was impressed by Avahi’s approach to working with a start-up company. “As is the case with most technology start-ups, our platform plans changed quickly as we were working with Avahi,” says Yung. “But Avahi realizes this often happens with start-ups, and they overcome this obstacle by communicating frequently. When we needed a new schema to sync with our changes, they jumped right in and made it happen.”
In addition to the platform change, Avahi also helped WittGen realize it needed to migrate the pre- processing function from on-premises machines to the AWS cloud—where it’s easier to access more RAM and scale other compute resources to format the data for analysis. The extra memory is needed because there is no special code for the pre-processing since the data sets submitted by researchers differ significantly, and the file sizes are often 50GB or more.
“Setting up the pre-processing function in the cloud was not easy, but Avahi managed to do it,” says Jung. “This is also an example of how they proactively recommend ways for us to improve our platform. Their approach differs from many partners who just sit back and only do what you ask them to do.” By partnering with Avahi, WittGen accelerated how quickly the machine learning platform for cancer research will go to market—expected sometime in 2023. “Providing a service sooner—that the healthcare industry can use to treat people with cancer—is the biggest win,” says Jung. “But from an operating standpoint, it also gives us a stronger position to secure more funding.”
Adds Rhie, “The security controls Avahi applied are also critical—helping us demonstrate our platform complies with HIPAA regulations. This is a must given the sensitive patient information we process. We appreciate Avahi’s expertise in ensuring AWS environments are secure and compliant to go along with their ability to design the infrastructure and code the machine learning backend.”
Jung and Rhie also realize that the way Avahi has designed the infrastructure and the machine learning backend will allow their application to run more reliably and faster for the researchers and clinicians using the system. And because Avahi created an infrastructure-as-code (IaC) environment, each time WittGen needs to add a new feature or apply a fix to the code, the change can be made quickly—with minimal or no downtime for the platform.
At the conclusion of the project, Avahi provided WittGen with the implementation code and explained how to maintain the infrastructure and make changes to the machine learning backend when necessary. “This is not something partners always offer,” says Rhie. “They sometimes hold onto the code, hoping you will need to keep coming back for more services. But Avahi has given us the option to handle things on our own.”
Looking ahead Jung and Rhie say they won’t hesitate to reach out to Avahi again if extensive development is needed on the platform or to tune the AWS environment as WittGen grows its customer base and user demand spikes. “Avahi is the ideal partner in that we don’t look at them as a partner,” says June. “They are an integral part of our team and want to us succeed as much as we do.”