complexa-design

Complexa Design Skill

Drive the full four-stage complexa design pipeline: generate (flow matching + search) -> filter (top-N by reward) -> evaluate (refold with AF2/RF3/ESMFold) -> analyze (success rate, FoldSeek/MMseqs diversity). Pick the right pipeline config for the design intent, validate the run upfront so the user does not discover a missing ckpt mid-folding, run it, and emit a replayable manifest + per-design success CSV.

What this skill enables

• Protein binder design for protein targets (AF2 reward + ColabDesign refold).
• Ligand binder design for small-molecule targets (RF3 reward + RF3 refold).
• AME motif scaffolding with ligand context (motif + ligand features, RF3).
• Search-based optimization: single-pass, best-of-n, beam-search, fk-steering, mcts.
• Refold backends: ColabDesign (AF2), RF3, Boltz2, ESMFold (fast iteration).
• Pass-rate + diversity analysis with per-result_type thresholds.

Step 1: Pre-flight

Always run the shared preflight before launching a design — generation needs the GPU and the right checkpoint, evaluation needs AF2/RF3 weights and tool binaries. Bail early if the host cannot run the chosen pipeline.

bash .claude/skills/_shared/scripts/preflight.sh

Read ./complexa_setup/preflight.json and bail if any of these are missing for the chosen pipeline:

• gpu.available: false -> all pipelines fail.
• gpu.vram_gb < 40 -> generation OOMs at default batch_size: 16; lower to 8.
• ckpts.complexa[.ckpt] -> required for protein binder.
• ckpts.complexa_ligand[.ckpt] -> required for ligand binder.
• ckpts.complexa_ame[.ckpt] -> required for AME.
• env.AF2_DIR missing -> protein binder generation AND eval both fail. AF2 is the af2folding search reward (loaded at generation time), not just the colabdesign refold backend — so a missing AF2_DIR crashes at sample 0, before any structure is produced. This is why complexa download --complexa alone is insufficient; you also need --all (which brings AF2).
• env.RF3_CKPT_PATH or env.RF3_EXEC_PATH missing -> ligand binder / AME generation reward and default eval (rf3_latest) fail.

If a ckpt is missing, point at complexa-setup and have the user run complexa download --complexa-<variant> first.

Step 2: Pick the pipeline

Complexa has one default pipeline (protein binder) and two extensions (ligand binder, AME / enzyme). The pipeline is selected entirely by the configs/search_*_pipeline.yaml you pass to complexa design — each YAML pins its own model checkpoint, autoencoder, targets dict, default reward, and default refold backend. Switching pipelines is just "swap the config path and the target name comes from a different dict".

Default — protein binder

complexa design configs/search_binder_local_pipeline.yaml \
    ++run_name=pdl1_v1 ++generation.task_name=02_PDL1

Use this when the user says "design a binder for X", "PDL1 binder", "de novo binder", "design proteins for a target", etc. — i.e. the target is a protein surface. The config pins complexa.ckpt + complexa_ae.ckpt, reads targets from configs/targets/targets_dict.yaml, rewards with AF2 (af2folding), inverse-folds with SolubleMPNN, and evaluates with ColabDesign / AF2. If the user did not specify, this is what they want.

Extension A — ligand binder (small-molecule pocket)

complexa design configs/search_ligand_binder_local_pipeline.yaml \
    ++run_name=v11_v1 ++generation.task_name=39_7V11_LIGAND \
    ++metric.binder_folding_method=rf3_latest

Switch to this when the user says "ligand binder", "small-molecule pocket", "SMILES target", "ATP-binding protein", or names a target ending in _LIGAND / from the FAD / SAM / OQO / 7V11 / etc. families. The config also activates LoRA (r=32, lora_alpha=64) which is required for the released ligand checkpoint — leave the lora: block alone.

Extension B — AME / motif + ligand (enzyme scaffolding)

complexa design configs/search_ame_local_pipeline.yaml \
    ++run_name=ame_chm ++generation.task_name=M0096_1chm

Switch to this when the user says "scaffold a motif near a ligand", "active-site design", "enzyme scaffolding", "AME", or names a target like M0024_1nzy, M0096_1chm (the M####_<pdb> AME task naming). The config sets env_vars.USE_V2_COMPLEXA_ARCH=True (the CLI runner injects it into the subprocess) and uses both MotifFeatures and LigandFeatures. Default search is single-pass; switch to best-of-n only if you also enable the CompositeRewardModel (commented out in ame_generate.yaml).

Pipeline cheat sheet — what changes when you switch

Knob	Protein binder (default)	Ligand binder	AME (enzyme)
Pipeline YAML	configs/search_binder_local_pipeline.yaml	configs/search_ligand_binder_local_pipeline.yaml	configs/search_ame_local_pipeline.yaml
Model ckpt	complexa.ckpt	complexa_ligand.ckpt	complexa_ame.ckpt
Autoencoder ckpt	complexa_ae.ckpt	complexa_ligand_ae.ckpt	complexa_ame_ae.ckpt
Targets dict	configs/targets/targets_dict.yaml	configs/targets/ligand_targets_dict.yaml	configs/design_tasks/ame_dict_v2.yaml
Task-name pattern	<NN>_<NAME> (e.g. 02_PDL1, 22_DerF21)	<NN>_<PDB>_LIGAND (e.g. 39_7V11_LIGAND)	M####_<pdb> (e.g. M0096_1chm)
USE_V2_COMPLEXA_ARCH	(unset → v1)	(unset → v1)	"True" (set in YAML)
LoRA	(none)	required (r=32, alpha=64)	required
Default search algo	best-of-n	best-of-n	single-pass
Reward model	AF2 (af2folding)	RF3 (rf3folding)	null (no reward at default)
Inverse folder	soluble_mpnn	ligand_mpnn	ligand_mpnn
Default refold backend	colabdesign (AF2)	rf3_latest	rf3_latest
Analysis result_type	protein_binder	ligand_binder	motif_ligand_binder
Required ckpts (complexa download)	--complexa --all (AF2 in community)	--complexa-ligand --all (RF3 in community)	--complexa-ame --all (RF3 in community)

For the full per-pipeline breakdown (reward weights, success thresholds, analysis modes), see reference/pipelines.md.

Step 3: Gather parameters

Use AskUserQuestion to fill in the four parameters that vary every run. Default to sensible production settings if the user has no preference.

• Target name — must be a key in the relevant dict (targets_dict.yaml for protein binder, ligand_targets_dict.yaml for ligand, ame_dict_v2.yaml for AME). Confirm it exists before running — do not let the pipeline fail with "Unknown target" after generation starts. Check membership directly:

  complexa target show <name>                       # prints the entry, or errors if absent
  # or grep the dict:  rg '^\s*<name>:' configs/targets/targets_dict.yaml
  

  If the target is not in the dict, hand off to complexa-target first, then come back here — do not stall on "Unknown target". The clean chain is:

  1. Detect the miss (e.g. user asks for "EGFR" but complexa target show EGFR errors).
  2. Switch to the complexa-target skill: gather the PDB source, chain/residue spec (or ligand SMILES), hotspots, and binder length, then register the entry.
  3. Verify with complexa target show <name> and complexa validate target <config> --target <name>.
  4. Return to this skill at Step 4 with the now-valid target name.

• Run name — a short identifier appended to the output dir (e.g. pdl1_v1).

• Search algorithm — default to beam-search with beam_width=8 and n_branch=4 for production. Use single-pass for a quick smoke test.

• Evaluation refold backend — protein binder defaults to colabdesign (AF2); ligand/AME default to rf3_latest. Use esmfold for fast iteration (worse but seconds per sample). ESMFold weights ship inside complexa download --all, but that sub-download needs HF_TOKEN set in .env or it silently rate-limits and leaves no ESMFold dir — confirm community_models/ckpts/ESMFold exists (complexa download --status) before relying on the esmfold backend.

Step 4: Validate

Validate before running. This is cheap (seconds) and catches missing ckpts, missing reward/eval weights, and a missing target entry — any of which would otherwise abort the pipeline mid-evaluation after hours of generation.

complexa validate design does NOT accept ++ Hydra overrides. The validate subcommand only takes <type> <config> [--target NAME]; passing ++generation.task_name=… makes it exit with "unrecognized arguments". It also reads the config YAML as written — overrides you plan to pass to complexa design are not applied during validation. So validate the bare config, and use --target (only available on validate target) to check a target other than the config's own generation.task_name:

# Validates ckpts + reward/eval weights, and the target named *in the config*.
complexa validate design configs/search_binder_local_pipeline.yaml

# To check a specific target (e.g. one you will pass via ++generation.task_name),
# validate it explicitly — this path DOES check dict membership:
complexa validate target configs/search_binder_local_pipeline.yaml --target 02_PDL1

The validator returns non-zero on errors and prints a pass/fail report. Re-run until it returns clean.

What validate design does not cover (verify these yourself):

• Override-dependent settings — e.g. if you will pass ++metric.binder_folding_method=esmfold, validation still checks whatever the config's default backend is. Confirm your overrides match the weights you downloaded.
• The target you pass at run time via ++generation.task_name= — validate it separately with validate target … --target <name> (above), or confirm it is in the dict directly (complexa target show <name>).

Step 5: Run the pipeline

complexa design is the right tool for the full 4-stage run — it orchestrates generate → filter → evaluate → analyze as sequential subprocesses with a shared run name, log dir, and multi-GPU split (see run_design_pipeline in src/proteinfoundation/cli/cli_runner.py). Re-implementing that manually loses the per-stage log routing and progress prints.

Use ++ (forced) Hydra overrides; they apply to all stages. The minimal production protein-binder invocation:

complexa design configs/search_binder_local_pipeline.yaml \
    ++run_name=pdl1_v1 \
    ++generation.task_name=02_PDL1 \
    ++generation.search.algorithm=beam-search \
    ++generation.search.beam_search.beam_width=8 \
    ++metric.binder_folding_method=colabdesign

For ligand binder / AME, swap the pipeline YAML and target name per Step 2's cheat sheet — every other override above is pipeline-agnostic and can be reused as-is.

Add --verbose to stream logs to the terminal instead of ./logs/. The skill does not poll progress — the user re-invokes if they want a status; point them at complexa status and ./logs/design_pipeline_*/.

Direct module invocation (debug fallback)

To debug a single stage without pipeline orchestration, invoke the underlying Hydra module directly. complexa generate CONFIG is just a logged subprocess wrapper around this:

python -m proteinfoundation.generate \
    --config-path "$(realpath configs)" \
    --config-name search_binder_local_pipeline \
    ++run_name=debug_pdl1 \
    ++generation.task_name=02_PDL1

Same pattern for proteinfoundation.{filter,evaluate,analyze}. Use this when you want to attach ipdb, run under nsys, or skip the pipeline log dir. Prefer complexa generate/filter/evaluate/analyze for normal one-shot runs (you get logging and parallel job splitting for free).

AME-specific gotcha (when running AME with RF3 refold): RF3 will try to complete missing atoms on the ligand based on its CCD code, which produces shape errors in RMSD calculations and the wrong structure. Before RF3 sees the PDB, rename the ligand residue to L:0 so RF3 treats it as a generic ligand and skips atom completion. If your AME generation outputs already encode the ligand this way (the canonical Complexa pipeline does), no extra step is needed; otherwise patch each PDB with atomworks.io:

from atomworks.io import load_any, to_pdb_file
atom_array = load_any("my_design.pdb")[0]
ligand_mask = atom_array.chain_id == "A"
atom_array.res_name[ligand_mask] = "L:0"
to_pdb_file(atom_array, "my_design_rf3_ready.pdb")

Same applies if you're piping AME outputs into the complexa-evaluate-pdbs skill with an RF3 backend. Skip the rename for AF2/colabdesign refold or for non-AME pipelines.

Wall-clock at default (nsteps=400, beam_width=8, batch_size=16, 100 designs, colabdesign eval) is ~30–120 minutes on a single A100/H100.

Step 6: Collect results

Outputs land in two directories. Surface both:

ls ./inference/${CONFIG_STEM}_${TASK}_*${RUN_NAME}/   # generated PDBs + filter
ls ./evaluation_results/${RUN_NAME}/                  # per-design CSV + analysis

Read the combined results CSV and summarize:

ls ./evaluation_results/*/binder_results_*_combined.csv
ls ./evaluation_results/*/motif_binder_results_*_combined.csv  # AME
ls ./evaluation_results/*/res_filter_*_pass_*.csv              # success rate
ls ./evaluation_results/*/res_div_foldseek_*.csv               # FoldSeek diversity

Pull the success rate from res_filter_binder_pass_*.csv, the per-design metrics (interface pAE, pLDDT, scRMSD) from the combined CSV, and FoldSeek TM-score diversity from res_div_foldseek_*.csv.

Empty diversity ≠ zero diversity. FoldSeek diversity is only computed when the foldseek binary is installed (FOLDSEEK_EXEC in .env). If it is missing, the analysis step silently emits the diversity column with empty values — it looks like a computed result but is a tool-missing failure. Before reporting diversity, confirm tools.foldseek.exists: true in ./complexa_setup/preflight.json. If it is false, tell the user diversity was not computed (install foldseek and re-run complexa analyze), rather than reporting "0 diversity".

Report top-N designs by i_pAE (protein binder) or min_ipAE (ligand binder).

Step 7: Emit manifest

Drop a JSON manifest beside the results so the run is replayable. The shared helper captures the command, config, git SHA, and pointers to the result CSVs.

python3 .claude/skills/_shared/scripts/write_manifest.py \
    --output-dir ./evaluation_results/${RUN_NAME} \
    --command "complexa design configs/search_binder_local_pipeline.yaml ++run_name=${RUN_NAME} ++generation.task_name=${TASK}" \
    --skill complexa-design \
    --out ./run_manifest.json

Surface the manifest path and the result CSV to the user.

Most-common overrides

The 10 overrides that cover ~90% of runs. Full reference (every key, type, default) is in reference/overrides.md.

Override	Default	What it controls
++generation.task_name=<name>	(per config)	Which target / motif task to design for
++run_name=<str>	(config stem)	Output dir suffix and CSV tag
++generation.search.algorithm=beam-search	best-of-n (binder/ligand), single-pass (AME)	Search strategy
++generation.search.beam_search.beam_width=8	4	Beam-search width (more = better designs, slower)
++generation.args.nsteps=200	400	Diffusion steps (fewer = faster, lower quality)
++generation.dataloader.batch_size=8	16	Drop to 8 on a 40GB GPU
++generation.filter.filter_samples_limit=500	1000	Top-N samples to keep after filtering
++metric.binder_folding_method=esmfold	colabdesign (binder), rf3_latest (ligand/AME)	Evaluation refold backend
++metric.num_redesign_seqs=8	2	ProteinMPNN/LigandMPNN/SolubleMPNN sequences per design
++aggregation.success_thresholds.i_pAE.threshold=10.0	7.0 (protein binder)	Loosen / tighten success criteria

Hardware requirements

Resource	Minimum	Recommended
GPU	1x CUDA GPU, 40 GB VRAM	A100 / H100 / L40S, 80 GB VRAM
CPUs	16	24 (the ncpus_ default in every pipeline config)
Disk	50 GB at ./inference/ + ./evaluation_results/	200 GB for sweep runs
RAM	32 GB	64 GB+

Typical wall-clock for 100 designs, beam_width=8, default nsteps=400:

• Protein binder + colabdesign refold: ~60–120 min on 1x A100/H100.
• Ligand binder + RF3 refold: ~90–180 min (RF3 dominates).
• AME + RF3 refold: ~120–240 min.
• Any pipeline + ESMFold refold: ~30–60 min (fast iteration).

Bumping gen_njobs=2 and eval_njobs=2 halves wall-clock on a 2-GPU host. See .claude/skills/_shared/reference/hardware.md for per-pipeline VRAM tables.

Troubleshooting (common cases)

Symptom	Cause	Fix
CUDA out of memory in generate	batch_size: 16 too big on 40GB GPU	++generation.dataloader.batch_size=8
CUDA out of memory in evaluate	AF2 / RF3 batched too aggressively	++eval_njobs=1 and ++metric.num_redesign_seqs=2
InterpolationKeyError: AF2_DIR	colabdesign eval but .env does not set AF2_DIR	Set AF2_DIR in .env or ++metric.binder_folding_method=esmfold
InterpolationKeyError: RF3_CKPT_PATH	RF3 eval but RF3 not installed	complexa download --all or switch eval backend
KeyError: 'task_name' not in target_dict_cfg	Target not in targets_dict.yaml / ligand_targets_dict.yaml / ame_dict_v2.yaml	Use complexa-target skill to add it
0 designs pass success thresholds	Defaults too strict for this target	Loosen via ++aggregation.success_thresholds.*

For the full list (chain-ID mismatches, hotspot residues, ligand residue renaming for RF3, missing inverse-folding models, etc.) see reference/troubleshooting.md.

---

For per-pipeline details (which model, reward, inverse folding, evaluation backend, result_type, LoRA settings, USE_V2_COMPLEXA_ARCH toggle), see reference/pipelines.md.

For the full override reference (every generation.*, metric.*, aggregation.* key with type, default, example, and effect), see reference/overrides.md.

For all troubleshooting cases (cause + fix + source), see reference/troubleshooting.md.