Ranomics
Computational protein structure visualization on a research workstation
Service

AI protein binder design services

De novo protein binders designed computationally with RFdiffusion, BindCraft, and Boltzgen, then validated experimentally through display screening

The limits of traditional binder discovery

Library-based approaches explore a tiny fraction of sequence space and depend on immunization or prior binders as a starting point. For novel targets or challenging epitopes, the starting library is often the bottleneck. You screen what you have, not what you need.

What generative protein design changes

Diffusion-based protein design models generate novel binder scaffolds from scratch, guided by target structure and hotspot constraints. Combined with experimental validation, this collapses the discovery timeline from years to weeks and removes the dependence on pre-existing libraries.

How it works

From target structure to confirmed binders

Seven stages from structural input to experimentally validated binder sequences. The same team runs every stage — no handoff between compute and wet lab.

01

Target preparation

You provide a target structure (PDB, AlphaFold model, or cryo-EM density). We identify binding hotspots and define the epitope region computationally, using interface energy decomposition and evolutionary conservation analysis.

02

De novo design

RFdiffusion, BindCraft, and Boltzgen generate backbone scaffolds conditioned on the target hotspot. ProteinMPNN designs sequences for each scaffold. Boltz-2, ESMFold, and ColabFold provide structural validation of top candidates.

03

In silico filtering

Candidates are filtered by predicted binding confidence, structural plausibility, expressibility, and manufacturability heuristics. Typical campaigns reduce 50,000 designs to 500-2,000 for synthesis.

04

Synthesis and display

Filtered sequences are synthesized as a pooled library and displayed on yeast or mammalian cell surfaces using Ranomics' established display platforms, coupling genotype to phenotype for every candidate.

05

Screening and selection

FACS or MACS selection against labelled target. Stringency is tuned to your affinity threshold. Multiple selection rounds can be run to enrich for tight binders.

06

NGS hit calling

Sorted populations are sequenced by NGS. Enriched clones are ranked and delivered as a prioritized hit list with binding data and sequence-level resolution.

07

Deliverables

Sequence-verified binder hits with binding evidence, ready for your downstream expression and characterization pipeline. Final report includes ranked sequences, display data, and NGS analysis.

What we can design

Binder scaffold formats

Scaffold selection is driven by target geometry, epitope accessibility, and downstream application requirements.

Nanobodies (VHH)

Single-domain antibody fragments. Compact, thermostable, and ideal for intracellular targets, enzymes, and GPCRs. Well-established expression in yeast and E. coli. The most versatile scaffold for de novo design campaigns.

Miniproteins

Ultra-compact binders (<60 amino acids) designed to engage concave epitopes, receptor pockets, or protein-protein interfaces. High stability and low immunogenicity potential. Strongest for geometrically constrained binding sites.

Custom scaffold binders

Target-specific scaffolds when existing frameworks are suboptimal. Scaffold selection is guided by target geometry, epitope topology, and your functional requirements. Includes designed repeat proteins, cystine knots, and other non-antibody formats.

Why Ranomics

What sets this AI binder design program apart

Most computational design services stop at predicted sequences. We validate every design experimentally before it reaches you. The same team runs the GPU compute and the wet lab — no handoff, no translation loss.

In-house GPU compute

We run design campaigns on cloud GPU instances (H100/A100) provisioned on-demand. Campaign scale: 10,000-60,000 designs per run. No external API quotas or shared batch queues.

No handoff between compute and wet lab

The same team that designs your binder validates it. Computational output feeds directly into our display screening platform, eliminating translation loss between stages.

Hotspot-guided design

We condition our generative models on experimentally defined or computationally predicted hotspot residues. This produces higher hit rates versus unconstrained design across the full target surface.

NGS-first analysis pipeline

Hit calling is done by sequencing enriched populations — quantitative, reproducible, and sequence-resolved from the first round. No guesswork from colony picking.

Structural inputs

Target inputs we work with

We accept multiple structural formats. Confidence filtering is applied where structural quality is uncertain.

Input type
Notes
Crystal structure (PDB)
Preferred. Enables precise hotspot definition with high geometric accuracy.
AlphaFold2 / AF3 model
Accepted. Confidence filtering applied using pLDDT and PAE thresholds.
Cryo-EM density map
Case-by-case. Contact us to discuss resolution requirements and preprocessing.
Homology model
Accepted with caveats on binding geometry. We flag uncertainty in hotspot definitions.
Applications

Use cases for de novo binder design

De novo binder design is particularly strong for targets where no prior binder exists or where conventional library approaches have failed.

Therapeutic antibody alternatives (nanobodies, miniproteins)
Diagnostic capture reagents
Protein-protein interaction inhibitors
Intracellular target engagement
Biosensor development
Research tool binders for novel targets
Agricultural biologics (crop protection, plant growth factors)
Animal health therapeutics and diagnostics

Have a target? Let's design your binder.

Send us your target structure and binding requirements. We will assess feasibility and propose a design campaign within 5 business days.

FAQ

AI binder design questions

What do I need to provide to start a project? +

At minimum: a target structure or high-confidence model, a defined epitope or functional region of interest, and your downstream application context. We will guide you through the rest in a scoping call.

Do you design antibodies or only nanobodies? +

Our primary de novo scaffolds are nanobodies (VHH) and miniproteins. For full-length antibody discovery, we recommend pairing with our yeast display platform. Contact us to discuss what format best fits your application.

How many design candidates do you screen? +

Typical campaigns screen 500-5,000 computationally filtered sequences in the display assay. Campaign scale — design size, number of selection rounds — is defined in your statement of work and depends on target difficulty and your affinity requirements.

Can you work with confidential target information? +

Yes. We operate under mutual NDA. All project data — target structures, sequence files, screening results — is kept strictly confidential and is not shared or used for any purpose beyond your project.

What is the typical project timeline? +

Our fixed-scope AI Binder Sprint runs 6-8 weeks from target intake to confirmed hits. Custom campaigns with extended design rounds or additional screening can take longer. Timeline is specified in the SOW at project start.

Ready to design your protein binder?

Tell us about your target and application. We will assess feasibility and propose a design campaign.

Start a project →