Biotech and healthcare companies around the world are conducting critical research to discover new therapies and treatments for a variety of diseases, which is even more important in the midst of a global pandemic.

Goldfinch Bio is one of those companies, focusing on discovering and delivering precision therapies for patients with kidney disease. Like many companies, Goldfinch Bio receives an influx of data that comes from varying sources which then needs to be analyzed, curated, and shared in order to inform clinical trials. Goldfinch translates these discoveries into new therapies that target the molecular causes of kidney diseases and, in doing so, helps create new, innovative treatment options for patients.

We recently hosted a talk with Adam Tebbe, Senior Director of Informatics and IT at Goldfinch Bio, to learn more about how data and technology are helping Goldfinch advance its mission, and to discuss how the challenges many companies are facing today with data will inform the future of biotech advancements.

As background, Rocket has helped Goldfinch to automate the processing of large scale genomic and longitudinal patient data to simplify their complex workflows and create efficient data visualizations. In short, we’ve built software to help their researchers to do better work, faster.

You can check out the full talk via the video below, or read on for a high-level summary.

Kidney disease by the numbers

Chronic kidney disease, or CKD, causes more deaths than breast cancer or prostate cancer. It is the under-recognized public health crisis, affecting an estimated 37 million people in the U.S. (more than 1 in 7 adults). 1 in 3 American adults is at risk for CKD.

Only 23,400 Americans received a kidney transplant in 2019, while about 100,000 Americans are still waiting for one. Less than one-third of these transplants came from living donors. Living and deceased kidney donors are crucial: 12 people die every day while waiting for a kidney transplant. While dialysis and transplants are still very much the standard care for most patients, new treatments are desperately needed to help this under-served patient population.

Enter Goldfinch Bio.

Goldfinch has been collecting samples from affected patients, sequencing them, and enrolling them into a Kidney Genome Atlas (KGA) to try to build a data set large enough that the company can use statistical methods to identify potential drug targets with genetic evidence.

This was an ambitious endeavor, and Goldfinch is a small company with very limited engineering staff. So they decided to bring in our team to help.

Building an MVP

The vision was to build a dynamic web-based application where Goldfinch could surface all of the data the company was collecting and producing as part of the KGA to help optimize target identification, patient stratification, and the exploration of the relationship between genotype and phenotype.

There was an aggressive timeline to launch an MVP, but Rocket was up for the challenge. We started with a design sprint and moved quickly into development. We quickly got to work to build a prototype of the platform so we could begin testing and iterating based on feedback from the research team. We learned that the interconnectedness of the platform was enabling the researchers and the biologists to take data, examine it, and be confident in what they’ve found, all in a much more efficient way. This is crucial, as it allows the team to move faster in pursuing which gene mutations they want to pursue further research on.

Not only did we produce anMVP on time and on budget, but we have continued the partnership adding a tremendous set of features and capabilities to the application since its launch.

The tech stack

[Tech folks, this section is for you]

What we’ve built for Goldfinch consists of a React SPA served directly from Amazon S3. This talks to an API built on a mostly serverless stack in AWS including Lambda, API Gateway, and Cognito for auth. There are three distinct environments for dev, staging, and production of the application. All of the code is tested and deployed using a CI pipeline built on GitLab.

Backing the API is an event-driven pipeline that produces everything the app needs on-demand.

The only persistent server in the stack is a relational database. The rest of the back-end is a carefully designed data lake that persists data across numerous different data stores including RDS Aurora and Athena. It’s not only very operationally efficient but cost-efficient as well.  

When someone uploads a new study from the app the pipeline is kicked off. The data presented to users is derived from petabytes of raw data. This data is processed, refined, and distilled to a much smaller, more manageable data set. This data is supplemented, enriched, and annotated with public data to try to give the end-users everything they need to better understand the genetic drivers of kidney disease and identify potential new drug targets that will help these patients that so desperately need new treatments.

Engineering and product design

The Goldfinch platform is built around the patient. Because of this, it’s increasingly important that the company has a strong engineering team helping to build out the infrastructure to manage, analyze and share all of the data and insights that drive our target identification pipeline. Sitting on top of this is product design, which in scientific domains is often overlooked. But for Goldfinch, it was important that the product designed was a professional software application, and looked that way as well. Thinking through the workflows they wanted to optimize, designing the interfaces, and studying user interaction were all important aspects to what was being built.

The future

Goldfinch’s vision is to eradicate kidney disease through precision medicine. The ability to evaluate large amounts of data, automate processes with technology, and ultimately speed up their workflow is a big deal in reaching that vision.

Data is being generated at an exponential rate, and this won’t change anytime soon. But with the help of Rocket, Goldfinch is saving time by doing less manual work, and they are able to easily see complex data visualizations through the platform that would have previously required third-party tools and manual efforts to pull together. These visualizations allow the team to unearth important patterns that will inform further research and ultimately treatment in the future.

Going forward, Goldfinch is discussing new ways the company can do a better job of interacting with the patient community and getting them engaged with the KGA. Having a software platform that is end-user focused will be absolutely critical in achieving this, and we are glad to have played a role in creating that platform for Goldfinch.