protein-design-skills

Run BioEmu (Biomolecular Emulator) — Microsoft Research's deep generative model that samples from the **approximated equilibrium (Boltzmann) distribution** of structures for a protein monomer, given its amino acid sequence. BioEmu is **not** a single-structure predictor: it's the AF3-class "many predictors" diffusion model whose output is an **ensemble** (an `.xtc` trajectory of backbone frames) approximating MD-style equilibrium, at orders-of-magnitude lower cost than running MD. Use this skill when: (1) Sampling a **conformational ensemble** of a protein monomer from its sequence (the headline use case — chignolin, folded/unfolded coexistence, fast-folder benchmarks, etc.), (2) Mapping **conformational changes** relevant to function — formation of **cryptic pockets**, local **unfolding** events, large-scale **domain rearrangements**, (3) Predicting **folding free energies / protein stability** (ΔG_fold) directly from the sampled ensemble (the v1.1 and v1.2 checkpoints were trained wi

Run BindCraft — a hallucination-based de novo binder design pipeline that couples **AF2 backpropagation** (ColabDesign), **ProteinMPNN** redesign, **AF2 reprediction**, and a full **PyRosetta** interface-score filter pass against a single target PDB. One `bindcraft.py` (or `sbatch bindcraft.slurm`) call runs an endless trajectory loop until the requested number of designs that pass every filter is reached. Use this skill when: (1) De novo designing **mini-protein binders** (~65–150 aa) against a protein target — the canonical BindCraft use case, (2) Designing **peptide binders** (≤ ~25 aa) using the dedicated `peptide_3stage_multimer` advanced preset + `peptide_filters`, (3) Designing for a **β-sheet-rich** target or a target where the binder *should* be β-sheet (`betasheet_4stage_multimer` preset), (4) Selecting a **hotspot patch** (`target_hotspot_residues`: `"56"`, `"A1-10,B1-20"`, `"A"`, or `null`) to focus AF2 on a specific epitope, (5) Producing AF2-validated designs ranked by **interfac

Run **CAVER 3.0 / 3.01** — the canonical tool from the Loschmidt Laboratories / Masaryk University for **identification, geometric characterization, clustering, and visualization of transport pathways (tunnels, channels, pores)** in macromolecular structures. CAVER operates on PDB files (a single static structure or an ensemble of MD snapshots) and returns the tunnels leading from a user-specified starting point inside the protein to the bulk solvent. Use this skill when: (1) Finding the access tunnel(s) connecting a buried active site / cofactor / co-substrate / catalytic residue to the protein surface, with their bottleneck radius, length, curvature, throughput, and priority — the headline CAVER use case, (2) Analyzing tunnel **dynamics** across a molecular-dynamics trajectory (10 / 100 / 1 000+ snapshots), clustering equivalent tunnels across frames into a small number of stable channels and ranking them by `Priority` (averaged throughput), (3) Comparing tunnels across **multiple ho

Run the **Biohub `esm` repository** (formerly EvolutionaryScale) — the world model of protein biology that ships **ESMC** (state-of-the-art protein language model), **ESMFold2** (AF3-class structure prediction with proteins + DNA + RNA + ligands), **ESM3** (generative model over sequence / structure / function), and **ESMC Sparse Autoencoders** (interpretable feature decomposition of ESMC's internal representations). Use this skill when: (1) Producing **per-residue or pooled embeddings** of one or many sequences with ESMC (300M / 600M / 6B) for downstream classification, regression, retrieval, clustering, or representation learning, (2) **Zero-shot mutation scoring / pseudo-perplexity / entropy** on a sequence by sampling masked logits (the headline ESMC use case for variant-effect prediction), (3) **Fine-tuning** ESMC for a task (PEFT / LoRA classification or regression head) or doing a **layer sweep** to find the best feature layer for a downstream probe, (4) Predicting **all-atom 3D

Use Biotite — a fast, NumPy-backed Python library for computational molecular biology — to read, manipulate, analyze, and write biomolecular structures and sequences, and to fetch data from biological databases. This is the toolkit/"glue" library of the collection: it does not fold or design proteins, it lets you operate on the structures and sequences that the predictors/designers consume and produce. Use this skill when: (1) Parsing, editing, or writing **structure files** — PDB, mmCIF / PDBx, BinaryCIF, MOL / SDF, GRO, and MD trajectories (XTC / TRR / DCD / NetCDF), (2) Selecting / filtering atoms, residues, or chains with NumPy boolean masks over annotation arrays (the `AtomArray` / `AtomArrayStack` data model), (3) **Superimposing** structures and computing **RMSD / RMSF / lDDT / TM-score** — e.g. comparing a designed backbone to its AlphaFold3 / Boltz / Chai prediction, or two predictions to each other, (4) Structural analysis: **SASA**, **secondary-structure** assignment, **hydrogen

Run Protenix — ByteDance's open-source, trainable reproduction of AlphaFold 3 for high-accuracy biomolecular structure prediction. Protenix folds complexes of proteins, DNA, RNA, small-molecule ligands, and ions in one pass, with post-translational/nucleotide modifications, covalent bonds, MSAs, templates, and optional pocket/contact constraints. Use this skill when: (1) Predicting the 3D structure of a **protein / nucleic-acid / ligand complex** from sequences + ligand specs (the headline use case), (2) Folding a **single protein** (monomer or homo-/hetero-oligomer via `count` / `id`) with an MMseqs2 MSA searched automatically, (3) Modeling **protein–ligand** binding where the ligand is a CCD code, a SMILES string, or a 3D structure file (SDF/MOL/MOL2/PDB), (4) Predicting **protein–DNA / protein–RNA** complexes, optionally with RNA MSA, modified bases, and double-stranded DNA (two complementary strands), (5) Adding **covalent bonds** (e.g. glycosylation, covalent inhibitors, cyclic peptid

Run PLACER (Protein-Ligand Atomistic Conformational Ensemble Resolver, formerly ChemNet) — a Baker-lab graph neural network that operates entirely at the atomic level to denoise / rebuild small-molecule and protein-sidechain atom positions, and to generate stochastic ensembles that model conformational heterogeneity. Use this skill when: (1) Docking a ligand that is **already present** in an input PDB / mmCIF into its binding pocket and scoring pose confidence (the headline use case), (2) Predicting / rebuilding protein **side-chain** conformations around a ligand or in an apo pocket (sidechain repacking with confidence), (3) Generating a conformational **ensemble** (50-200 stochastic samples) of a ligand + pocket to study heterogeneity rather than a single pose, (4) Validating designed enzyme / heme-binder / metalloprotein active sites by checking whether PLACER re-resolves the intended ligand pose at low prmsd, (5) Co-predicting **multiple ligands** in one pocket (`--predict_multi`), kee

Run the FAIR `fair-esm` package — the original Meta Fundamental AI Research reference implementation of the ESM family of protein language and structure models. Use this skill when: (1) Extracting per-residue or per-sequence embeddings from a protein language model (ESM-2 6 variants from 8M → 15B, ESM-1b, ESM-1v, ESM-MSA), (2) Predicting protein 3D structure end-to-end from a single sequence with **ESMFold** (`esm-fold` CLI or `model.infer_pdb()`), (3) Designing sequences for a fixed backbone — fixed-backbone (a.k.a. inverse-folding) sequence design with **ESM-IF1** (`GVPTransformer`, single-chain or multi-chain complex), (4) Scoring conditional log-likelihoods of candidate sequences against a backbone (variant ranking, design scoring, perplexity), (5) Zero-shot variant effect prediction on deep mutational scans with **ESM-1v** ensembles or ESM-MSA (wt-marginals, masked-marginals, pseudo-perplexity), (6) Unsupervised contact prediction from attention maps (`return_contacts=True

Build inverse-rotamer active-site assemblies from a Rosetta matcher/enzdes constraint (CST) file using PyRosetta. Use this skill when: (1) Constructing theozyme / active-site stubs (catalytic sidechains placed around a ligand) as starting points for de novo enzyme design, (2) Preparing inputs for **RFdiffusion All-Atom** (RFdiffusionAA) enzyme design pipelines — outputs include `REMARK 666` enzdes records so they drop straight into the published heme-binder-diffusion workflow, (3) Exhaustively enumerating clash-free combinations of catalytic rotamers + small idealized helix/strand backbone stubs around a small-molecule substrate / cofactor, (4) Hosting one catalytic residue inside an externally-provided **motif PDB** (e.g. a CYS loop from a cytochrome P450) while still enumerating the other CST residues as inverse rotamers, (5) Filtering rotamers by Dunbrack cumulative probability (per-CST or global), per-CST secondary structure, and per-CST random subsampling to control co

Build composable protein design pipelines with ProtFlow — a Python framework that wraps RFdiffusion, RFdiffusion3, LigandMPNN/ProteinMPNN/SolubleMPNN, ESMFold, ColabFold (AF2), AlphaFold3, Boltz, Caliby, Frame2Seq, ProteinGenerator, PLACER, AttnPacker, SigmaDock, Rosetta, GROMACS, DSSP, FPocket, Propka, TM-align, RMSD/biopython metrics, and more behind a single Poses + JobStarter + Runner API, with first-class SLURM array-job and local execution. Use this skill when: (1) Composing multi-step design pipelines (generate backbone → design sequence → predict structure → score → filter → iterate) without writing bespoke cluster-management code. (2) Running any of the supported tools on a SLURM cluster as array jobs, or locally as parallel subprocesses, behind a unified interface. (3) Managing a pandas-DataFrame-backed `Poses` object that carries all structures, sequences, scores, residue selections, and motifs through the pipeline, persisted as JSON/CSV/Pickle/Parquet/Feather. (4) Filtering / r

Run BoltzGen, an open all-atom diffusion model for universal binder design — generates ranked, filtered binder candidates from a single YAML specification. Use this skill when: (1) De novo designing protein, peptide, cyclic-peptide, helicon, or stapled-peptide binders against a protein target, (2) Designing nanobody / VHH CDR loops against a target, (3) Designing antibody (Fab) CDR loops against a target, (4) De novo designing protein binders against a small-molecule ligand (SMILES or CCD), with built-in binding-affinity prediction, (5) Designing protein binders against DNA / RNA / nucleic-acid targets (e.g., de novo zinc fingers against DNA), (6) Designing against disordered regions or peptides shown as a fixed sequence with no structure, (7) Redesigning or optimizing residues on an existing complex (`protein-redesign` protocol — symmetric dimers, scaffold optimization), (8) Inverse-folding a fully specified backbone with BoltzGen's IF model (`--only_inverse_fold`) — replaces ProteinM

Run Boltz-1 / Boltz-2, MIT-licensed open foundation models for biomolecular structure and binding-affinity prediction. Use this skill when: (1) Predicting structures of protein monomers, multimers, and protein-protein complexes, (2) Predicting protein-ligand complexes (ligands as SMILES or CCD codes), (3) Predicting protein-DNA, protein-RNA, or mixed nucleic-acid complexes, (4) Predicting binding affinity (probability of binding and log10(IC50)) between a small molecule and a protein target — Boltz-2 only, (5) Validating designed binders / antibodies against a target (alternative to AlphaFold2 / Chai-1 with comparable or better accuracy on complexes, and 1000x faster than physics-based FEP for affinity), (6) Folding with pocket / contact / covalent-bond / template restraints, (7) Modeling cyclic peptides, modified residues (CCD), and non-canonical covalent bonds, (8) Running batched inference across one or many GPUs from a directory of YAML inputs. Covers installation (pip, GitHub, Do

Run DISCO (DIffusion for Sequence-structure CO-design), a multimodal generative model that jointly co-designs a protein sequence and 3D structure, conditioned on (and co-folded with) arbitrary biomolecules — small-molecule ligands, ions, metals, DNA, or RNA. Use this skill when: (1) Generating de novo protein sequence + backbone from scratch (unconditional, length-controlled), (2) Designing a protein binder / scaffold around a small-molecule ligand provided as a SMILES string, an SDF / MOL / MOL2 / PDB file, or a PDB CCD code (single-component or multi-component glycans), (3) Designing proteins around metals, ions, organometallic catalysts, or reactive intermediates (e.g. heme enzymes for new-to-nature carbene transfer, B–H insertion, C(sp³)–H insertion, cyclopropanation), (4) Designing RNA-binding proteins conditioned on a fixed RNA sequence, (5) Designing DNA-binding proteins conditioned on a single-strand or explicitly-paired double-strand DNA sequence, (6) Partial-mask / sequence-r

Run Chai-1, a multi-modal foundation model for molecular structure prediction. Use this skill when: (1) Predicting the structure of a protein, protein-protein, or protein-peptide complex, (2) Predicting protein-ligand complexes (ligands provided as SMILES or CCD codes), (3) Predicting protein-DNA, protein-RNA, or mixed nucleic-acid complexes, (4) Predicting glycosylated proteins or proteins with non-canonical covalent modifications, (5) Validating designed binders / antibodies against a target, (6) Folding with user-supplied contact / pocket restraints to guide complex assembly, (7) Generating MSAs via the ColabFold server or supplying your own `.aligned.pqt` files / templates, (8) Running batch inference distributed across multiple GPUs. Covers installation, the `chai-lab` CLI (`fold`, `fold-batch`, `a3m-to-pqt`, `citation`), the Python API (`run_inference`, `run_folding_on_context`), Chai's FASTA-like input format (proteins, ligands, glycans, modified residues, DNA, RNA), the restraints

Use and extend the Foundry toolkit (RFdiffusion3, RosettaFold3, ProteinMPNN/LigandMPNN/SolubleMPNN/EnhancedMPNN/ThermalMPNN). Covers CLI/Python APIs, checkpoint management, common design workflows, and — most importantly — how to extend the toolkit (registering new MPNN checkpoints, adding new MPNN architectures, plugging new inference engines into the Hydra configs). Trigger when a user mentions foundry, rf3, rfd3, mpnn (any variant), enhancedmpnn, thermalmpnn, the `foundry` CLI, FOUNDRY_CHECKPOINT_DIRS, or asks to add/register a new MPNN model/checkpoint.

Run de-novo protein design with Genie 3 — a fast, all-atom, SE(3)-equivariant diffusion model. Use this skill when: (1) Generating unconditional monomeric protein backbones (length 50–600+), (2) Solving motif-scaffolding problems (single- and multi-motif), (3) Designing protein binders for a target with hotspot residues, (4) Running beam search or iterative-conditioning binder pipelines, (5) Loading legacy Genie 2 (Cα-trace) checkpoints, (6) Training the Genie 3 model from scratch. This skill covers installation, the unified `genie3` CLI (run / generate / evaluate / status / train), the experiment YAML schema, the three application workflows, multi-GPU + multi-node sharding, problem-set preparation, output layout, and the in-silico success filters. Pairs with: `proteinmpnn` (inverse folding), `colabfold` / `esmfold` (structure prediction back-ends used by the evaluation pipeline), `foldseek` (cluster successful designs), `protein-qc` (QC thresholds), and `binder-design` (tool selection between Genie 3, BindCr