/Article
At Adaptyv Bio, we are on a mission to enable protein designers to focus on doing what they do best — designing proteins — while our automated lab handles the experimental characterization of their designs.
One year ago, when we first publicly launched Adaptyv Bio, we ran a survey among protein designers to understand what lab workflows they need and what currently blocks them from iterating on their designs faster. We got an amazing amount of feedback and in almost all cases, the number one bottleneck was better access to experimental data. So we're trying to fix that.
With Adaptyv Bio you can now generate lab data for your protein designs with just a few clicks. Choose your experiment type, upload your designs and you’re ready to go! Our lab will automatically assemble the DNA sequences, synthesize the proteins, run the assays and process the data.
Want to see how it works? Watch the video below!
Want to see some data? Check out this case study where we benchmark RFdiffusion designs!
Protein designers should be designing, not pipetting
Proteins are great. We all are literally made from proteins and those tiny molecular machines power everything we do, from digesting food to moving our muscles to fighting off viruses. Countless industries rely on engineered proteins, such as antibody therapeutics, gene therapy vectors and gene editing enzymes as well as proteins used in food production, cosmetics and recycling. Thanks to advances in AI models like AlphaFold, RFdiffusion & co, computationally designing new proteins has turned from science fiction into reality. But a protein design on the computer only gets you half way. While the models have been progressing rapidly, the lab infrastructure to actually test those proteins experimentally hasn’t advanced and is just not set up to respond to the speed and data demands of this new age of AI-driven protein design.
This means that protein engineering companies can’t just focus on designing proteins but until now they also need to build and manage their own wet lab to test those designs. That is not only expensive but also distracting - the skills required to design proteins using ML models are very different from the ones needed to design biological assays and run lab infrastructure.
At Adaptyv Bio, our goal is to empower thousands of new protein designers by providing the necessary wet lab infrastructure so they can focus on designing, rather than pipetting.
Think about how developers can focus on building products and use managed infrastructure like AWS, Vercel, Modal and co to handle hosting, deployment or inference for them. In the early days, software companies still had to maintain their own servers and infrastructure. It was only when the field grew large enough that there was an opportunity for companies to specialize in providing infrastructure services. This infrastructure in turn makes it easier for developers to build apps and services, some of which can become infrastructure themselves for new applications (infrastructure-app-cycle).
We believe protein engineering has now reached a similar pivotal moment. With AI models for protein engineering gaining momentum, there's a massive demand from protein designers for better wet lab capacity to generate experimental data about their designs.
By building a foundry instead of just yet another internal lab, we can benefit from economies of scale for lab automation, negotiate lower prices with suppliers, have tighter ecosystem integration and build better experimental characterization tools.
This way we can make it cheaper, faster and less frustrating for protein design teams to get the experimental results they need. They can save themselves the cost and the headache of running a wet lab and instead they can dynamically scale their data generation using our platform, cutting down timelines to get to a good protein design for their application.
Get started designing proteins using Adaptyv Bio
We think that there should be more protein designers so we want to make it extremely easy to get started getting data about proteins. No endless Teams meetings, Typeforms to fill out or opaque ‘request a quote’ pages - our production assays are available in our Experiment Configurator with transparent pricing information.
When you confirm your first experiment we’ll give you access to our Foundry Portal where you can see all your experiments at a glance, launch new ones and get notifications about status updates while we run the assays and analyze your data.
- Dashboard Overview: The main dashboard provides a snapshot of all your ongoing projects, categorized by status: Planned, In Progress, and Completed.
- Experiment Status Updates: Each experiment's status is updated in real-time so you’ll always know what’s going on in the lab and when your results are ready.
- Results Visualization: Your data is displayed with interactive charts and graphs to help you understand easily how each of your designs performed in the lab.
- Download Data: You can download all data generated from your experiments to build your own figures and train your AI models.
We also know how messy biology can be and how hard it is to separate signal from noise in experimental data. This is why our platform automatically optimizes assay conditions for you and we track various confidence metrics, including all the metadata for all results.
For this release we’ve focused on streamlining everything around protein binding assays. This unlocks a vast number of applications, from engineering therapeutic antibodies to novel miniproteins to replacing the thousands of animal-derived research antibodies with synthetic alternatives.
- Binding Screening: This is the fastest method to differentiate hits from non-binders and allows you to rapidly identify promising candidates from a large pool of designs. It’s particularly useful for screening de novo designs generated from AI models like RFDiffusion.
- Binding Affinity Characterization: Here, we run multiple rounds of binding kinetics measurements to generate high-quality affinity data for your designs. With this workflow you can dial in the exact binding strength that you need for your application and compare performance between different designs.
- Epitope Binning: For designs that have shown binding activity, our epitope binning workflow can determine if proteins compete for the same binding site on your target. This way you can understand the specificity and potential cross-reactivity of your designs.
More to come
We have exciting plans for the coming months - if you’re interested in working with us on any of the following, just get in touch via hi@adaptyvbio.com
- Expanding foundry capabilities: We are continuously working to expand the range of assays available on our platform as well as increasing throughput and bringing down cycle time. If you are a protein designer, let us know which assay types and product features you need. Whether it's more data analysis tools, API access, integration of protein property prediction, or streamlined licensing options for the proteins you design, just shoot us your ideas via email!
- Model benchmarking and training: We are running more benchmarking campaigns for protein designs generated by RFDiffusion & co. If you are building new models and want to generate or license large scale training or validation data, get in touch with us!
- Protein design challenges: Foundations and grant makers can partner with us to set up bounties to harness the creativity of the protein design community for solving protein design problems. We’d love to set up regular protein design challenges where protein designers can compete to achieve the highest lab-validated performance for a specific application.