Genoplan establishes a foundation for Asia's leading genetic testing service on AWS

Genoplan chose Amazon Web Services (AWS) as its cloud provider, as AWS is linked with the experimental data analysis system environment. Genoplan has changed its IT structure, determining that a cloud-native architecture is the right tool to digest B2C data.

Genoplan has created a web pipeline for B2C services, as well as a data pipeline design that connects the lab to legacy and cloud operations. The dual data pipeline design was critical in being able to satisfy consumer expectations to receive reports as soon as possible. Delivering reports within 10 business days is the service level of leading global genetic analysis companies, which in effect means analyzing experimental data very rapidly.

Regarding Genoplan's strategy, Ahn Chong Hyun, data science team leader, says, “With the more than 700,000 genes we analyze, we face tremendous time constraints to view and verify. Hence, we automated part of the verification process using algorithms, and we’ve built the data pipeline in such a way that the researchers could inspect it manually to check if there was an error.”

Genoplan aims to reduce the time needed for DNA test data verification and establish a serverless data processing environment. Determining the need for an easily scalable database to store its rapidly growing data reliably, Genoplan chose Amazon DynamoDB, which can quickly query and process experimental data, and selected AWS Lambda, AWS Step Functions, and AWS Batch to complete the automation-based pipeline. In anticipation of a future rapid rise in data volume, intake, and processing, Genoplan is considering processing data using a data lake.

Once the experimental data is generated, this information goes up in the pipeline to Amazon S3 and is stored the moment it is triggered. Experimental data verification is an automated process based on algorithms, and only erroneous information is checked by the human eye. When the final data comes out, it is converted and entered into Amazon DynamoDB. All of this processing is done by AWS Lambda, AWS Step Functions, and AWS Batch in a serverless environment. This flow is both economical and practical. If there are a large number of genetic test requests, Genoplan can quickly provision resources to AWS Lambda and process those requests simultaneously, automated in Amazon SNS. When the number of test requests drops, it can likewise scale down its resources.

Genoplan's web services are focused on availability and security. Team leader Kim Bum Jin says, “We initially built a basic web service and there was a potential problem with ensuring high performance availability and security. That's why we changed the database to Amazon RDS, which we installed on the fly. We're considering introducing CDN services and web firewalls to keep pace with the Asian market.” he adds.

Genoplan's B2C service is gaining market traction  Regarding Genoplan’s strategy, CEO Kang Byung Kyu says, “With the introduction of microarray sequencing in 2019, we have begun serving our individual customers, and requests are growing more than 20% per month. Genoplan will focus on developing and analyzing DNA chips used for Asians, and will actively expand to Korea, Japan, Singapore, and Southeast Asia. Due to the nature of the business, as the number of users increases, the amount of data increases exponentially, thus we expect to expand our business with stability through AWS. We also use the cloud as a pay-as-you-go plan, so we can transparently predict the cost of expanding our different business stages, which will help us plan better.” 

The contribution of machine learning to the data science process is one of the core competencies of Genoplan, and seen as something indispensable. Hyun says, “We are dealing with data in two ways: The first is data analysis, which is collected for research purposes. The second is to spot or prevent errors in the experiment process. This method requires a lot of trial and error in order to establish a data analytics model through machine learning, which is less costly to do in the cloud."

"To further reduce the burden, we plan to use Amazon SageMaker to build a testing model and to test existing models, which currently rely on our in-house GPU server. We plan to use a fully managed service and deploy the complete model used for analysis," Hyun added.

Meanwhile, Genoplan expects cloud services like AWS to be a catalyst for the company’s success in the full genome sequencing era. Kang says, “In the 1990s, it cost about KRW 100 billion over the span of 10 years to analyze a single genome. The cost of performing full sequencing to analyze 3 billion sequences has now dropped to several hundred thousand Korean won. In the future, this cost will be further reduced due to the advancement of cloud technology.”

He added “We have made significant reductions in the manual effort required in verifying DNA experiment data. This allows us to fulfill our commitment to consumers to deliver their test results within 10 business days. Even if customer test kit analysis requests swell, consumers can receive reports of results within 10 days with the near-unlimited scalability provided by AWS.”