| Structure prediction |
| AlphaFold 2Predict a complex with ColabFold AlphaFold2, the standard for binder-target interfaces. | 2for one structure | Run |
| Boltz-1Legacy Boltz-2 folding path; prefer BOLTZ2 unless you need its template-mapping inputs. | 1for one structure | Run |
| Boltz-2Co-fold proteins with ligands, DNA, RNA, templates and pocket restraints; can also predict binding affinity. | 2for one structure | Run |
| Chai-1Chai-1 folding of protein monomers and complexes, from Chai Discovery. | 2for one structure | Run |
| ESMFoldFold a single protein chain from sequence alone. The cheapest way to fold. | 1for one structure | Run |
| ESMFold 2Fold proteins, complexes, and DNA/RNA/ligand systems with the full ESMFold2 model. | 2for one structure | Run |
| ESMFold 2, capsidFold one capsid protein and assemble an AAV 60-subunit icosahedral capsid. | 2for one structure | Run |
| ESMFold 2, fastFold single sequences with the small ESMFold2 checkpoint, for screening throughput. | 2for one structure | Run |
| Mol-1Mol-1 full biomolecular co-folding for proteins, nucleic acids, ligands and ions. | 2for one structure | Run |
| OpenDDEAlphaFold3-class co-folding from Aureka, with an antibody-antigen specific checkpoint. | 2for one structure | Run |
| OpenFold 3OpenFold3 co-folding for explicitly specified multi-entity complexes. | 2for one structure | Run |
| ProtenixAlphaFold3-class co-folding from ByteDance, strong on antibody and VHH complexes. | 2for one structure | Run |
| Binder design |
| BindCraft, PyRosetta-freeHunt for de novo miniprotein binders by AlphaFold2 hallucination, without PyRosetta filters. | 10per accepted design | Run |
| BoltzGenEnd-to-end BoltzGen design: diffusion, inverse folding, refolding, ranking and filtering in one job. | 10per 100 designs | Run |
| BoltzGen, design and inverse foldBoltzGen diffusion and inverse folding, then Phi folding and interface metrics. | 15per 100 designs | Run |
| BoltzGen, design and inverse fold onlyBoltzGen diffusion and inverse folding, stopping at sequences. | 11per 100 designs | Run |
| BoltzGen, design onlyBoltzGen diffusion only, stopping at backbones. | 10per 100 designs | Run |
| BoltzGen, fullBoltzGen design followed by Phi's own AlphaFold2 and interface-metric scoring. | 13per 100 designs | Run |
| Dyno PsiGenerate binder backbones with Dyno's DynoPSI-1 flow-matching model. | 10per 100 designs | Run |
| ESMFold2-DesignOptimise a binder sequence against a target by gradient guidance through ESMFold2. | 5a run | Run |
| Genie3Generate protein backbones with Genie3: unconditional sampling, motif scaffolding, or target-conditioned binder design. | 15per 100 designs | Run |
| GerminalHunt for epitope-targeted VHH designs by hallucination, co-folding and antibody-aware sequence design. | 10per accepted design | Run |
| Proteina-ComplexaCo-design protein-binder sequences and all-atom complexes with Proteina-Complexa. | 10per 100 designs | Run |
| PXDesignGenerate all-atom binder backbones with ByteDance PXDesign-d. | 10per 100 designs | Run |
| RFantibodyDesign antibody and VHH CDR loops against an antigen epitope with RFdiffusion-Ab. | 10per 100 designs | Run |
| RFD4-ProteinaCo-design structure and sequence with RFD4-Proteina: protein binders, antibody/VHH CDRs, motif scaffolds, partial-diffusion refinement and inverse folding in one model. | 45per design | Run |
| RFdiffusionGenerate poly-glycine binder backbones with the original RFdiffusion. | 10per 100 designs | Run |
| RFdiffusion 3Generate all-atom binder backbones against a target with RFdiffusion3. | 10per 100 designs | Run |
| Sequence design |
| ESM-IF1Inverse-fold a backbone with ESM-IF1: design sequences for one chain, or score existing sequences against the backbone. A GPU alternative to PROTEINMPNN. | 1per 32 backbones | Run |
| Evo 2, generateGenerate DNA continuations from prompts with Evo 2. | 1a run | Run |
| PottsMPNNDesign sequences for a backbone with PottsMPNN, or score existing sequences against one; ProteinMPNN's contract on KeatingLab's Potts-head sampler. | 1per 32 backbones | Run |
| ProteinMPNNDesign sequences for a backbone, or score existing sequences against one. | 1per 32 backbones | Run |
| Proton-PottsMPNNProtonation-aware sequence design on CPU: pH-switch design over a combined_lambda stability/selectivity sweep, or per-residue protonation labelling. | 1per 32 backbones | Run |
| Analysis |
| DockQCompare a predicted complex against a native reference with DockQ. | 1a run | Run |
| Fold and score noveltyFold sequences with ESMFold2-Fast, then score their structural novelty. | 2for one structure | Run |
| MaSIF-sitePredict which surface patches of a target are likely interaction sites. | 1a run | Run |
| MSA PairformerPredict zero-shot effects for one or more variants against a reference sequence. | 1a run | Run |
| Phi ScoreScore predicted complexes with Phi's trained affinity model. | 1a run | Run |
| Seq2SymmClassify a protein sequence into its homo-oligomer symmetry (C1, C2, D2, ...). | 1a run | Run |
| Structural noveltyScore how structurally novel a design is against PDB100, AFDB50 and CATH50. | 1a run | Run |
| Pocket discovery |
| LacunaFind cryptic binding pockets by generating and analysing a conformer ensemble. | 1a run | Run |
| Embeddings |
| ESM CProduce ESM-C embeddings and masked-LM logits for sequences. | 1a run | Run |
| Evo 2Score and embed DNA sequences with Arc Institute's Evo 2 genome language model. | 1a run | Run |