Pacesa Lab
Biomolecular design • De novo Biology • AI Therapy
BindCraft2: the next generation of de novo protein binder design
Out now

BindCraft2

BindCraft2 is the next generation of our platform for user-friendly de novo protein binder design. Like the original, it was built to make binder design easy and accessible to non-experts, without requiring high-throughput screening. It is far more computationally-efficient and unlocks a range of design modalities, such as miniproteins, peptides, VHHs, and antibodies, together with powerful multi-target optimisation and therapeutically-relevant design objectives.
Preprint (coming soon)Try it online ↗GitHub ↗Wiki ↗BindCraft1 paper ↗

Using BindCraft2 we want to push computational design toward a future where engineering biomolecular interactions becomes a routine skill in every biologist's toolbox, with standardised protocols and outputs that can be used both in basic research and therapeutic applications.

Why make a sequel

The first BindCraft made de novo binder design accessible to biological labs without computational expertise and screening pipelines. BindCraft2 builds on that, but also removes the barriers that kept many projects out of reach, such as compute cost, runtime, and a narrow range of targets and binder formats.

◌
Democratise binder design for every lab, not just specialists
◌
Automated and easy to use, with minimal human input
◌
Optimized for experimental success, not just hacking in silico scores
◌
Open source and free for academic and industry use (except hosting services)
◌
Designed to slot into both discovery research and therapeutic pipelines
BindCraft2 can design large rigid binder modalities for structural biology and imaging applications.

One pipeline, many possibilities

The real strength of BindCraft2 is its versatility. One automated pipeline covers a wide span of target types, binder formats and downstream goals, so you rarely need a different tool for a different job.

◈
Inputs: structured domains, disordered regions, multiple targets at once, steer away from off-targets, and make molecular glues across targets
◈
Outputs: miniproteins, linear peptides, cyclic peptides, large binders, VHHs, antibodies, ankyrin repeat proteins, homooligomers, multidomain binders
◈
Induced fit accounted for on both the target and the binder side
◈
Design against a specific conformational state, for example the active form of a receptor, while detargeting from the inactive form
◈
Research reagents and probes for detection, pulldowns and pathway perturbation
◈
Therapeutic leads in miniprotein and antibody-fragment formats
◈
Large rigid binders that add mass to small targets for cryo-EM visualisation
◈
A precise way to probe, regulate, and disrupt biological pathways
BindCraft2 can design binders even against disordered targets.

Optimise your binder's properties

Every campaign can be tuned with optional objectives that go beyond making a binder stick. Steer it onto a difficult epitope, humanise it for the clinic, harden it against proteases or lock in rigidity, and when a target is stubborn, BindCraft2 adapts on its own to still find candidates.

◍
Forced targeting: pin binders onto stubborn, suboptimal epitopes
◍
Humanisation: bias toward human-germline sequence with fewer immune liabilities
◍
Protease resistance and designed disulfides for tougher, more rigid binders
◍
Termini controls to keep chain ends free for tags, fusions and grafting
◍
Desperation mode: keep iterating settings on hard target sites to achieve binding
◍
Mix and match the objectives your experiment actually needs
A de novo designed scFv-format binder in BindCraft2.

Design modalities

One automated pipeline, many binder formats. Pick the one that matches your target and your downstream application.
De novo miniproteins

De novo miniproteins

Compact single domain binders with high expresion rates and stability. The most reliable format for lab reagents, biosensors and therapeutic leads.

Large binders

Large binders

Bigger, rigid binders that add defined mass and recognisable features to a small target for cryo-EM, or that cover large, flat epitopes.

Linear & cyclic peptides

Linear & cyclic peptides

Short peptides that can thread into grooves and pockets, or head-to-tail or disulfide-linked macrocycles with higher rigidity and less entropy.

Homo-oligomers

Homo-oligomers

Symmetric, multivalent binders for avidity, receptor clustering or agonism, target multimerisation, and the natural fit for symmetric targets.

Multidomain binders

Multidomain binders

Two (or more) binding domains on a single chain, for multiparatopic or multispecific reach across several epitopes or different targets.

VHH (single-domain antibody)

VHH (single-domain antibody)

Single-domain antibodies that reach into concave and cryptic epitopes and enzyme active sites. Clinically derisked modality but trickier to design.

scFv & Fab

scFv & Fab

Classic antibody-fragment formats for the design of therapeutic antibodies, designed using human-germline frameworks. Very difficult modality.

Ankyrin repeat proteins (ARP)

Ankyrin repeat proteins (ARP)

Non-antibody, disulfide-free, high-stability scaffolds of human origin, that support intracellular use and easy multivalent fusions.

Induced-fit & fold-switch

Binders that change shape, or even their whole fold, on binding, for conformation-selective, allosteric and switchable sensor designs.

Targets and inputs

BindCraft2 works from many kinds of target inputs, and its multi-target optimisation is where it stands apart: design one binder against several proteins, across species, or against a divergent target while steering clear of its paralogs.
Structured target

Structured target

A folded domain from a PDB or CIF file. Pick the chains, name the epitope you care about, or let BindCraft2 find an appropriate site.

Disordered region or motif

Disordered region or motif

An intrinsically disordered region or linear peptide given as a sequence. BindCraft2 generates binders to stabilise motifs that have no fixed structure.

Cross-reactive, multi-species

Cross-reactive, multi-species

One binder optimised to engage the same target across species, for example human and mouse, from a single design campaign.

Multi-specific, several proteins

Multi-specific, several proteins

One binder jointly optimised against several divergent proteins at once, a single molecule that recognises multiple partners.

Detargeting for specificity

Detargeting for specificity

Actively repel chosen off-targets so a binder hits its intended target and misses close paralogs or known off-targets from previous campaigns.

Molecular glue and bridging

Molecular glue and bridging

Engage multiple targets together to hold them in proximity in a defined geometry, for induced-proximity and glue-style designs.

What it still can't do

BindCraft2 still only deals with pure protein-protein interactions. Main limitations include:

  • No accounting for ligands (PTMs, small molecule, nucleic acids, etc)
  • Only natural amino acids, in both the binder and the target
  • Ranking reflects the probability of binding, not predicted affinity
  • Some target sites are hard and may still yield few passing designs
  • Target and binder size remain limited by your GPU memory
  • Still expect to screen 2-20 designs in the wet lab

Publications that experimentally validated BindCraft designs

BindCraft2
MainComing soon

BindCraft2

Preprint in preparation

Coming soon
One-shot design of functional protein binders with BindCraft
MainAugust 27 2025

One-shot design of functional protein binders with BindCraft

Pacesa M*, Nickel L*, Schellhaas C*, Schmidt J, Pyatova K, Kissling L, Barendse P, Choudhury J, Kapoor S, Alcaraz-Serna A, Cho Y, Ghamary KH, Vinue L, Yachnin BJ, Wollacott AM, Buckley S, Westphal AH, Lindhoud S, Georgeon S, Goverde CA, Hatzopoulos GN, Gonczy P, Muller YD, Schwank G, Swarts DC, Vecchio AJ, Schneider BL, Ovchinnikov S, Correia BE

Nature
Induced Estrogen Receptor SUMOylation drives SERD activity
BC2 BindersJune 26 2026

Induced Estrogen Receptor SUMOylation drives SERD activity

Hinterndorfer M, Schätz C, Schmitt S, Schönlein M, Hoi DM, Krecioch I, Frommelt F, Shlei M, Kater L, Pacesa M, Munoz M, Kempf G, Kladnik K, Batty P, Imrichova H, Aguirre JD, Högler S, Cavadini S, Seruggia D, Correia BE, Obenauf AC, Thomä NH, Winter GE

bioRxiv
Latent Generative Search unlocks de novo Design of Untapped Biomolecular Interactions at Scale
BC1 BindersMarch 16 2026

Latent Generative Search unlocks de novo Design of Untapped Biomolecular Interactions at Scale

Didi K, Reidenbach D, Penner M, Ravichandran S, Case M, Nichols M, Swanson E, Reis A, Prescott M, Qian Y, Qian D, Yang J, Li W, Li L, Shonai D, Gay S, Mallik BB, Chim HY, Chen L, Juantay MA, Klein H, Macintyre AU, Secor M, Granata D, Cao Z, Zhou G, Geffner T, Chen X, Livne M, Zhang Z, Zhang T, Gion K, Bronstein MM, Steinegger M, Deibler K, Soderling S, Khmelinskaia A, Hollfelder F, Dallago C, Kucukbenli E, Vahdat A, Ogden P, Kreis K

Whitepaper
Crowdsourced Protein Design: Lessons From the Adaptyv EGFR Binder Competition
BC1 BindersApril 24 2025

Crowdsourced Protein Design: Lessons From the Adaptyv EGFR Binder Competition

Cotet T-S, Krawczuk I, Stocco F, Ferruz N, Gitter A, Kurumida Y, Machado LA, Paesani F, Calia CN, Challacombe CA, Haas N, Qamar A, Correia BE, Pacesa M, Nickel L, Subr K, Castorina LV, Campbell MJ, Ferragu C, Kidger P, Hallee L, Wood CW, Stam MJ, Klounis T, Unal SM, Belot E, Naka A, Adaptyv Competition Organizers

bioRxiv