Boltz-2 Online: Co-Folding Protein-Ligand Complexes Without Local Setup
How to run Boltz-2 co-folding directly in the browser with ChemOrchestra — no Docker, no GPU setup, no command line.
Jul 18, 2026
Model
Boltz-2
Modes
General cofold + multi-ligand screening
Access
Browser-based, no local install
What co-folding does differently from separate fold-then-dock pipelines
The traditional pipeline predicts a protein's structure first, then docks a ligand into that fixed shape. Boltz-2 instead predicts the protein structure and the protein-ligand complex geometry jointly. For flexible or shallow binding pockets, that distinction matters: the pocket's shape can shift depending on what's bound to it, and a co-folding approach captures some of that induced-fit behavior instead of assuming the protein stays rigid throughout.
Boltz-2 General vs. Boltz-2 Screening in ChemOrchestra
ChemOrchestra exposes Boltz-2 as two workflow nodes. Boltz-2 General runs a single protein-ligand co-folding job — useful when you have one target and one candidate ligand you want a complex structure for. Boltz-2 Screening runs the same underlying model across multiple ligands against a target, letting you compare interactions across a candidate set in one pass rather than running each pairing separately.
What the confidence scores mean
Boltz-2's output includes per-residue confidence (pLDDT-style) and a predicted binding-affinity signal. Read both as a triage signal, not a final answer — use them to rank and filter a candidate list down to a manageable shortlist, then validate the top hits with orthogonal methods before committing synthesis resources. Treating a single co-folding score as proof of binding is the most common way teams over-trust this class of model.
From Boltz-2 output to your next experiment
A co-folded complex is a starting point, not an endpoint. Feed the predicted structure into ADMET prediction to check developability, or use it to inform where a linker or scaffold modification might improve binding geometry. Running Boltz-2 in the same workflow builder as the rest of your pipeline means the output is immediately usable downstream instead of living in a separate file you have to hand off.