We’re delighted to share an interview with Ben Kamens, CEO and Founder, Spring Discovery, whose AI software is making significant contributions to measuring cell behavior.
Please provide a brief overview of your company.
Spring gives AI superpowers to scientists, starting with the world’s best high content image analysis suite. We started back in 2017 with an emphasis on AI technology built just for scientists who use imaging for drug R&D, and we now offer this tech to others – making it easy for them to apply the latest neural networks coming out of Google and other research labs to their imaging experiments.
...our tools let scientists easily teach AI models to identify complex cell behavior in images just by visually clicking on cell phenotypes they’re interested in.
We’re obsessed with the idea of giving scientists new tools via great, usable, powerful, and fast software. For example, our tools let scientists easily teach AI models to identify complex cell behavior in images just by visually clicking on cell phenotypes they’re interested in. We want to put better tools in scientists’ hands so that they can learn from what are increasingly high-dimensional datasets for even the most complex biology, cell models, and assays. Scientists are trying to protect us from disease and we force them to use inferior software way, way, way too much.
Spring’s technology helps scientists rapidly drive to conclusions for phenotypic discovery – we’re used by partners around the world to identify top drugs and targets, get to mechanisms of action, distinguish between healthy/diseased and responder/nonresponder states, and more.
Spring’s technology helps scientists rapidly drive to conclusions for phenotypic discovery.
What is the customer need/opportunity that your company solves?
Any company that generates microscopy images has a bunch of problems to solve. They have a gold mine of valuable biological signals hiding in their data, but…
- It’s difficult to wield these enormous imaging datasets at even basic data infrastructure levels.
- Imaging data has unique and complex quality concerns that need to be expertly monitored with all sorts of assay bias detection that most tools are only beginning to fully appreciate.
- Assuming you can get past those roadblocks, it’s cumbersome, finicky, and time-consuming to try to measure even basic phenotypes across all your images, especially at a single-cell level –people spend all their time battling finicky segmentation settings and the like.
- Finally, we haven’t even mentioned the fact that bleeding-edge AI technologies open up a whole new route for interpreting biological signals in these images – most have barely started to scratch this surface.
Also, another bonus challenge customers face – all of the above gets dramatically more difficult when dealing with the most complex cell types like neurons/microglia/etc. or when dealing with co-cultures of many different cell types.
Spring’s AI tech helps teams do all of the above with a fast, scientist-friendly tool – so they can spend their time using the power hidden in phenotypic assays to identify top drugs and targets, deconvolute mechanism of action, figure out why certain samples are non responders, the list goes on and on.
How does your company solve these challenges?
Spring has an end-to-end cloud-based imaging analysis suite that takes customers from (1) easy ingestion of their data to (2) automatic AI-based segmentation and featurization of their images to (3) self-serve analysis to (4) easy export of reporting and data. This dramatically speeds up experimental analysis for even the biggest and most complex imaging datasets.
Scientists can also train their own AI models by visually clicking on images of cell phenotypes – a great example of how we use AI to unlock the expertise and intuition of human scientists.
Our product automatically generates both traditional features for all customer images (measuring things like cell shape, texture, intensity and the like) and purely AI embedding-based features for detecting the most meaningful features of cell phenotypes in a completely unbiased way. Scientists can also train their own AI models by visually clicking on images of cell phenotypes – a great example of how we use AI to unlock the expertise and intuition of human scientists. Combined, this lets scientists bridge the gap between unbiased AI-based analysis and interpretable, concrete features to use downstream in drug programs.
Thanks to the above, our technology excels at modeling complicated cell phenotypes and co-culture situations. Results are easily explorable and searchable at blazing fast speeds. All data is exportable into other tools in your ecosystem. And if you specialize in your own custom analysis metrics, you can upload those via our API and benefit from Spring’s frontend analysis architecture to expose your AI/data team’s work.
Like a labmate or copilot, Spring’s AI offers on-demand assistance to scientists, offering guidance towards phenotypic insights when requested, automatically detecting experimental bias and quality issues, and more. Finally, Spring’s software suite combines industry-standard traditional cellular image features with world-leading, unbiased AI-powered analysis so scientists can rank and sort drugs using multi-dimensional features, clustering, similarity/difference scores, and more.
Where have you seen success?
Our partners and customers use Spring’s technology for disentangling complex biology between healthy-vs-diseased states, ranking compounds in high-throughput screens, measuring nuanced functional differences between clinical candidates, understanding the difference between responder/non-responder samples, improving QC pipelines for cell therapies, and more.
Put simply, we get very strong responses from any group that generates lots of 2D imaging data and wants to get more insights out of it. We’re definitely a slam dunk for any group interested in or already using the cell painting assay. And for any group that also wants more exposure to the power of AI and how unbiased analysis can benefit their imaging data, our solutions are an easy way to dip toes into the cutting edge.
We are expanding beyond these initial customer segments, but those are our bread and butter right now.
What’s the benefit of joining the Tetra Partner Network?
TetraScience’s commitment to unrestricted access to scientific data is essential to move science forward. The promise of AI-driven scientific outcomes relies completely on having large volumes of data formatted properly to drive the AI model. We’re delighted to join the TetraPartner Network so that we can combine strengths to help scientists by providing significantly more data to feed our AI models, which is crucial to understanding complex biology.