GRASP Library Designer

Codon-optimize GRASP (Farley et al., NAR 2025) binder DNA for Golden Gate assembly.

PyPI: grasp-library-designer · Import: grasp_library

Open in Google Colab

Click a badge → run 0 · Install (PyPI) → fill the forms top to bottom. No GitHub token needed.

Notebook	Open
One-shot (one RNA → Golden Gate oligos for the binder gene)
Library (42-module redesign → GAP compile)

Direct links:

• One-shot: https://colab.research.google.com/github/JustABiologist/grasp-library-designer/blob/main/grasp_oneshot_designer.ipynb
• Library: https://colab.research.google.com/github/JustABiologist/grasp-library-designer/blob/main/grasp_library_designer.ipynb

Each notebook installs with:

%pip install -q -U "grasp-library-designer>=0.1.15"

Bundled GenBank modules, Potapov ligase-only matrices, and Pryor Golden Gate cycling matrices ship inside the package (materialize_project()).

Install locally

pip install grasp-library-designer
# optional notebook extras
pip install "grasp-library-designer[notebook]"

from grasp_library import materialize_project, build_default_config, LigationFidelityCalculator

project = materialize_project()  # ./grasp_library_project + GenBank
config = build_default_config(project / "input")
print(LigationFidelityCalculator(25, 18).set_fidelity(["AATG", "GATA"]))

From a blank Colab / Jupyter, you can also drop the Forms notebooks onto disk:

%pip install -q -U grasp-library-designer
from grasp_library import write_notebook
write_notebook("oneshot")   # or "library"
# then open the written .ipynb from the file browser

What each notebook does

One-shot does not emit combinatorial library modules. It builds an ORF that starts with ATG-Met, then the native solvating helix and PPR repeats, co-designs codon-aligned cut sites, high-fidelity 4-nt overhangs, and a synonymous CDS (codon optimality, cut-site depletion, synthesis heuristics), then wraps each fragment with inward-facing Type IIS arms and destination sticky ends. Design also writes a multi-record .gb (assembled ORF, ligated insert with destination overhangs, and each order oligo) labelled with CDS, PPR repeats, 4-nt overhangs, Type IIS cut sites, and the shared pool PCR primers.

Hard constraints (library path): the protein sequence is fixed and every movable four-base cut is restricted to the invariant ARELF motif. The search explores all motif-relative offsets 0–11 rather than only the four cut positions chosen in the paper. A candidate is therefore an (overhang, ARELF offset) pair, and each part is rematerialized before codon optimization. Objectives are ligation fidelity, codon optimality, and synthesis fitness.

Ligation fidelity is reported per physical six-overhang Level 0 reaction (and optionally as an explicitly labelled product across independently transformed blocks). The scalar is the orientation-invariant geometric mean of the two directional products. Stage-matched Pryor et al. 37↔16 °C Golden Gate cycling matrices are used by default: BsaI-HFv2 for Levels −1 and 1, and BbsI-HF (BpiI isoschizomer) for Level 0 redesign scoring. Potapov’s ligase-only data remain available as optional Level 0 surrogates; they contain no measured 16 °C matrix, so the program does not interpolate or blend static temperature matrices. These scores are optimization surrogates, not cloning guarantees. Synthesis QC distinguishes PASS, WARNING, and FAIL; vendor profiles remain transparent heuristics with vendor_acceptance_confirmed=False.

The order file contains double-stranded synthesis fragments with paired, inward-facing BsaI sites. The dashboard exposes exactly one physical 5′/3′ overhang pair for each cloning level. Every overhang is written 5′→3′. The deposited GRASP toolbox defaults are:

• Level −1: ACAT / ACAA.
• Level 0: CTCA / CTCG.
• Level 1: GGAG / AGCG.

The 3′ sticky ends are reverse-complemented internally when constructing the coding-oriented sequence. Thus the retained 3′ coding sites are TTGT, CGAG, and CGCT, respectively. Internal five-part and ARELF junctions are derived from the GRASP architecture rather than presented as extra dashboard overhang fields. When no custom acceptor sequence is provided, the exporter validates interface requirements but does not claim backbone simulation.

The exported GRASP tract is a PPR block set, not a standalone expression plasmid. The PPR block-chain check does not validate an entire Level 1 expression construct; promoter, upstream domain, effector, terminator, and acceptor context must be supplied separately.

For 14S and 19S, intermediate junctions are generated inside the invariant ARELF motif. They are architecture-derived and can be explored by the overhang redesign search without adding more dashboard fields.

Develop from source

git clone https://github.com/JustABiologist/grasp-library-designer.git
cd grasp-library-designer
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[notebook,dev]"

Package layout

grasp_library/                 # installable Python package
  data/profiles/.../genbank/   # bundled GRASP GenBank modules
  notebooks/                   # Colab Forms notebooks (also at repo root)
  paths.py                     # materialize_project()
  ...
third_party/dawdlib_golden_gate/   # Potapov ligation fidelity (AGPL)

License

AGPL-3.0 (required by the vendored GGAssembler / dawdlib ligation engine). See LICENSE and THIRD_PARTY_LICENSES.md.

GRASP sequences: Farley et al., Nucleic Acids Res. 2025.