Alloy is Building the World's First Fully Integrated Product Development
Organization - Read the full article
AI-DRIVEN ANTIBODY & NANOBODY DISCOVERY & ENGINEERING
From GenAI to measured molecules
We optimize generative protein-design models for your target product profile, then build and screen what they design. You get molecules and measurements, not software.
10,000+
up to 360°
EXPLORE ATX Reforge & ATX Cast
See how each program works
Learn how to get started, review results from real campaigns, understand what you receive, and explore the commitments behind every engagement.
HOW TO GET STARTED
Start with a molecule, or start with a target
The same discovery engine, lab, and deliverable, tailored to where your program begins.
Start with a Design Assessment
Whether you’re starting with a molecule or a target, the ATX Design Assessment helps define your priorities, evaluate the opportunity, and identify the right path into ATX-Reforge or ATX-Cast. Costs can be credited towards ATX-Reforge or ATX-Cast
Start with an antibody
Bring us a clinical asset, a shelved program, or a lead that needs rebalancing. You define the target product profile. We redesign, test, and rank new candidates against it.
Optimize the full profile
Rebalance affinity, potency, specificity, half-life, immunogenicity, and developability against the profile you define, rather than blindly trading one away for another.
Select with confidence
Come away with a ranked portfolio and the data behind every candidate in it.
Build a differentiated and future proof portfolio
Give your IP counsel a portfolio of candidates that are materially distinct at the sequence level from the parent you already own, each with measured properties and a documented design rationale.
Start with a target
A defined target, an epitope, a functional objective. No starting molecule. We steer our proprietary AI workflows to the epitope, build a focused library, and screen it.
Design for your epitope and function
Point the design at the target and epitope you define, with the functional objective written into the design brief before anything is built.
Find differentiated starting matter
Generate starting matter orthogonal to what immunization returns, for targets where conventional approaches keep producing the same handful of clones.
Carry hits forward into Optimization
Selected hits feed directly into an ATX-Reforge program, carrying the campaign data forward into optimization.
CAMPAIGN RESULTS
Performance across both starting points
Real campaigns, most of them run with a partner on their own molecule.
Optimization campaigns
In two design–build–test rounds, completed in approximately six weeks, candidates reached picomolar affinity, improved functional potency, and cross-species binding beyond the clinical comparator
Most leads reach picomolar affinity with zero developability flags within six weeks.
199/200
Designed variants bound
We expressed and characterized all 200 variants. 199 bound the target, 147 measured below 10 nM, including five in the picomolar range.
<100 pM
Affinity, reached in two rounds
Approximately six weeks of design, build, and testing produced sub-100 pM affinity alongside functional potency that exceeded the clinical comparator.
Human + cyno
Cross-reactivity not achieved by comparator
Designed using generative models alone, the molecule retained binding across both species through two optimization rounds while outperforming the clinical comparator on affinity and potency. Unlike the comparator, it does not require a surrogate for primate studies.
De novo design performance
Results from newer campaigns on targets for which only a predicted structure was available.
46/46
De novo designs expressed
All 46 designed molecules expressed, allowing the screen to evaluate binding performance rather than construct failure.
>10%
Top recombinant hit rate
More than one in ten designs produced a recombinant hit, in line with published de novo design benchmarks.
14 nM
Best first-pass binder
A fully de novo VHH reached 14 nM affinity in the first design round against a target with no experimental structure available.
CAMPAIGN DELIVERABLES
What arrives at the end of a campaign
A decision-ready package built for the scientists, program teams, and leadership who need to understand what worked, what changed, and what to do next
A denoising trajectory from a diffusion run. Successive steps, from noise to a folded single-domain backbone. Alloy has developed best-in-class algorithms for sampling candidate sequences from modern protein design models.
A ranked candidate portfolio
Candidates ordered against your target product profile, spanning performance and sequence diversity, so the trade-offs stay visible instead of averaging out.
The underlying data, in full
Every measurement behind the final designs: selection and enrichment data, expression, affinity, and developability readouts on the characterized set. Your scientists get the raw table, not a summary of it.
A one-page story per candidate
What changed, what improved, what it cost elsewhere, and how far the sequence sits from your parent molecule.
A sequence-space assessment
Design neighborhoods, edit-distance distributions, liabilities, and measured properties, laid out so your counsel can see which candidates sit furthest from the parent and exactly which data supports each one. Built to be handed to a patent attorney and to your engineering team without translation.
Comparative context, where we have enough of it
Where we have run enough comparable campaigns to say something statistical, we show you the distribution. Where we have not, we tell you the sample is too small.
A recommended next step, priced
Every report closes with what we would do next and what it would cost. Including when the recommendation is to stop.
OUR COMMITMENTS
What we put in writing
Three terms in every contract, settled before any work starts.
We stand behind the process
A defined, quality-controlled workflow with named remedies written into the contract. What the biology does is measured and shared with you as we find it.
The molecules you take forward stay yours
Option a candidate and both the molecule and its data belong to you. It stays out of our models and never reaches another partner. Ours improve on our own internal campaigns.
Give your leadership more evidence to work with
Candidates materially distinct at the sequence level, each with measured properties and its design record. Counsel gets filing options now, and a start on a next-generation version of your own molecule. Patentability and freedom to operate stay theirs.
getting started
Find the right path forward
In five business days, get a focused assessment of where to start and what to do next.
Bring us a molecule or a target, along with the outcome you want to achieve.
We evaluate the design opportunity, identify the key constraints, and recommend the right path forward.
If you choose to move forward, we turn the assessment into a design campaign and generate the molecules and experimental data to support your next decision.
