Platform

Enabling discovery and design of target-activating antibodies through large-scale activity measurement

Abalone Bio's FAST platform leverages engineered cells to evaluate hundreds of millions of antibodies simultaneously, assessing their ability to modify the activity of a target rather than merely binding to it. By producing extensive sequence-activity datasets, FAST not only uncovers rare activators for challenging targets like G-protein-coupled receptors but also enables their AI-powered design.

Lack of large-scale activity measurement was a major roadblock to target-activating antibodies

The lack of large-scale activity measurement has been a significant obstacle in creating target-activating antibodies, particularly for GPCRs and other membrane proteins. These targets present unique challenges.

Why are GPCRs and other membrane targets hard?

Proteins graphics

Proteins are hard to produce

Many in vitro discovery methods require large amounts of homogenous, properly folded protein. GPCRs are complex membrane proteins that are difficult to produce and purify.

Target-activating antibodies, particularly for GPCRs and other membrane proteins

Limited accessibility

GPCRs are embedded in cell membranes, exposing limited extracellular portions for antibody interaction.

GPCRs embedded in cell membranes

Complexity underlying activation

The lack of large-scale activity measurement has been a significant obstacle in creating target-activating antibodies, particularly for GPCRs and other membrane proteins. These targets present unique challenges:

Introducing FAST

our Functional Antibody Selection Technology platform

Measuring activity, not just binding, in high-throughput by harnessing “survival of the fittest” with engineered biology.

We harness Nature’s original algorithm

We have engineered cells to each measure the activity of a single antibody sequence variant - a massively parallel biological computer.

Engineering FAST cells to proliferate if carrying a functional antibody

We engineer each of FAST’s yeast cells- a synthetic biology workhorse eukaryotic organism similar to a simplified human cell- to produce a functional human target protein and a single antibody variant per cell.

Enabling discovery and creation of rare functional antibodies with large-scale functional datasets

We produce growth curves that reflect function for hundreds of millions of antibody sequences. These data are mined to discover rare hits in large antibody libraries, and to train protein LLMs to extrapolate into novel sequence space to create optimized hits.

Our FAST platform addresses fundamental challenges in antibody discovery and design, paving the way for breakthroughs for previously intractable targets.

Introducing FASTTM

our Functional Antibody Selection Technology platform

Measuring activity, not just binding, in high-throughput by harnessing “survival of the fittest” with engineered biology.

Five circular gauges with gradient progress bars indicating different levels on a dark background.

We harness Nature’s original algorithm

We test and identify highly functional individuals in large, diverse populations leveraging the survival of the fittest.
Abstract network of glowing blue nodes connected by lines on a dark background.

Engineering FAST cells to proliferate if carrying a functional antibody

We engineer each of FAST’s yeast cells- a synthetic biology workhorse eukaryotic organism similar to a simplified human cell- to produce a functional human target protein and a single antibody variant per cell.
Abstract representation of AI with connected lines and nodes on a dark background.

Enabling discovery and creation of rare functional antibodies with large-scale functional datasets

We produce growth curves that reflect function for hundreds of millions of antibody sequences. These data are mined to discover rare hits in large antibody libraries, and to train protein LLMs to extrapolate into novel sequence space to create optimized hits.

Our FAST platform addresses fundamental challenges in antibody discovery and design, paving the way for breakthroughs for previously intractable targets.

The conventional focus on affinity 
is flawed

Affinity
Activity

Our technology directly measures activity of all antibodies, eliminating biases from binding-based methods

Existing large-scale discovery technologies focus on attributes like binding affinity first, largely because that is historically an easy property to score and optimize.

Activity is tested very late in the discovery process on a few high-affinity antibodies targeting regions that may have nothing to do with altering the activity of the target. It’s like tightly grabbing a pair of scissors by the blades instead of more loosely by the handle- a high affinity but non-functional interaction.

Our large-scale activity datasets unlock the power of AI for activating antibody design

Other AI methods use static protein structures to attempt binding prediction, falling short for predicting activating binders for dynamic, large, complex membrane targets like GPCRs. FAST broadly samples the experimental landscape of antibody activity.

One FAST campaign generates activity datasets nearly 10X bigger than sequence datasets used to train class-leading antibody foundation models. By combining pre-trained antibody language models and our datasets, we enable model-enriched discovery and generative design of more active, developable antibodies.