Shiru: Accelerating lab protein discovery by 3x with Superblocks

Challenge: Analyzing protein testing results at scale 

Shiru is an Alameda, CA based company that leverages machine learning, bioinformatics, and precision fermentation to identify and create functional, sustainable ingredients for the plant-based industry. With demand for sustainable and functional food ingredients growing, Shiru is scaling its operations to identify new proteins at a rapid pace. 

‍

As part of its routine operations, Shiru's team must evaluate thousands of proteins, so internal tooling for data analysis and visualization as well as collaboration is critical for decision making by both lab scientists and company leadership. To enable this, Shiru's team had to solve the following challenges:

1. Data is siloed across multiple Postgres and Snowflake tables with no central UI, making it difficult for Scientists to self-serve and analyze testing results
2. Existing biotech R&D software has limited customization, which is needed for protein analysis visualizations and reports
3. As an R&D startup of Data Scientists, Biologists, and Research Scientists, building their own custom internal software from scratch is expensive, slow, and isn’t their core competency

Solution: 3x faster scientific analysis with Superblocks as the single pane of glass

With Superblocks, Shiru was able to build applications to track and understand high-throughput protein screening results across multiple dimensions.

1. Batch experiment results 

The batch experiment results application built in Superblocks allows lab scientists to quickly answer questions about protein expression by leveraging customized Plotly charts on top of measurement data in Benchling. With a single dropdown selection, users can see all protein candidates tested in a batch and an interactive distribution of results overlaid with granular metadata.

‍

2. Scoring proteins across experimental sets 

To identify the best candidates for further downstream validation and in-application testing, an aggregate view of all experiments displays proteins in an exportable table, where each row is clickable for deeper insights into protein characteristics, lab notes, and findings from experimentation.

‍

3. Protein inventory explorer

This internal application was built in just a few hours and joins data from multiple tables in Postgres into a single view.  Without running complex SQL queries, anyone in the company can now look up the protein inventory for location and status of various protein ingredients Shiru produces for both internal experimentation and external sampling.

‍

Why Superblocks: Turning Data Scientists into full-stack developers

1. Enabling Data Scientists to create UIs without React

As data practitioners, Shiru’s Data Scientists were already familiar with SQL and Python scripting to manipulate lab data, but not so much with creating entire UIs on top of this. Superblocks bridged the gap, enabling them to augment their tooling with frontend drag and drop components wired up to their backend Snowflake and Postgres.

2. Pandas and Plotly for complex data visualization

Support for writing Python directly in the editor made it easy for the team to incorporate libraries like Pandas and Plotly for merging data from multiple sources, as well as visualizing complex statistics with customized box charts that weren’t available with existing R&D software.

3. Ease of use for Builders and End Users

Having a platform that was user-friendly was paramount. Not only did this allow the scientists in the lab to iterate on tools quickly, but also made it simple for non-technical users to discover and gain insight into previously buried data, now surfaced in intuitive UIs.  

Results and future state 

With Superblocks, Shiru has adopted a data driven mindset, focusing on robust and scalable systems that deliver high fidelity data from the start, rather than manual and error prone processes common for smaller companies their size. This has decreased the time it takes to understand key results from screening experiments and move protein candidates through the testing pipeline. 

‍

As they continue to scale, Shiru plans to add new internal tools to correlate experiment data with millions more proteins, improve machine learning models, and find the best ingredients among a massive reservoir of candidates.