About Prism

Core principles

We are building the foundation to make protein motion measurable, predictable, and actionable.

Prism is building this foundation openly, empowering all scientists to use this data immediately to discover new biology.
Our Science

  • Building open, usable tools and methods

    We are openly building the methods, tools, and technologies to make protein motion measureable, actionable, and predictable. This will accelerate the ability for us to uncover fundamental principles of biology. Prism is fully open, in line with Astera’s Open Science Policy.

  • Creating new, integrated ways to do science

    We are a distributed team, combining full-time scientists at the Astera Institute with researchers in academic groups. Biological questions define our research areas. Capabilities define our technology areas. Our organizational structure, integrating scientists, engineers, and funders, allows us to which is what it takes to transform a field.

  • Answering pressing biological questions

    We are testing how conformational ensembles can answer questions that static structures cannot. We are learning to improve selectivity between closely related targets, to predict how a mutation shifts a protein's function, and to engineer new motion into proteins. Dynamics becomes something to design and not only to observe.
Why now?

Three shifts make now the moment to treat proteins as dynamic ensembles rather than static structures.

  • 01
    AI-driven structure prediction (such as AlphaFold) proved that biological insights invisible to us can emerge from quality data at scale.
  • 02
    Experimental methods around collecting ensembles are becoming faster, richer, and more routine.
  • 03
    Machine learning algorithms are starting to mature to learn from noisy, heterogeneous structural data.
Prism is making this transformation possible.
Why coordinated?

Progress is lost at the interfaces.

Our Science
A better method that the next tool cannot read is a method that stops there. A model trained on one group's data cannot be tested against another's. Fixing one area in isolation produces a better area and little else.
Only coordinated technology change, pairing with biological questions, will transform the field. Standards, formats, benchmarks, and infrastructure have to move together, because each one is what makes the next one usable.
Open questions

Making protein dynamics a well-posed AI problem.

  • How do dynamics matter most for answering biological questions?

    How do we use information about protein motion to understand evolution, drug design, evaluate disease variants, or engineer new proteins and cells?

  • What data should we collect?

    How do different structural measurements provide insights into function, and how do we evaluate this? We are leveraging AI and deep-dive biological questions to answer this question to guide scaling data collection.

  • What is the right benchmark for AI to predict protein dynamics?

    Structure prediction succeeded because the problem was well-defined, through years of community-led benchmarking. Ensembles have no agreed-upon target or benchmark. To take advantage of AI, we have to define and frame the prediction problem.
What's possible

What's possible if we achieve this goal?

Understanding dynamic conformational ensembles expands what we can ask of biology and what we can ultimately understand, design, and act on.

  • More precise drug design

    Design drugs against specific functional states, improving selectivity between closely related targets, and predictive control of protein function through alloster

  • Predicting disease mutations

    Predict how a mutation changes a protein's behavior rather than only its shape. Connect structural data to disease, where dysfunction often reflects a shift in dynamics.

  • Engineering protein dynamics

    Engineer new motion into proteins for biosensors, allosteric switches, enzymes, molecular logic, conditional therapeutics, and responsive materials. Dynamics becomes something to design, not only to observe.