Arpeggia

This is a port of the Arpeggio library to Rust, with a focus on identifying certain protein-protein interactions in PDB and mmCIF files.

Features

• Parse PDB and mmCIF files
• Parse user selection of chain groups
• Extract protein chains and residues
• Calculate distances between residues
• Identify protein-protein interactions
  • Steric clashes
  • VdW interactions
  • Hydrophobic interactions
  • Aromatic interactions
  • Cation-pi interactions
  • Ionic interactions
  • Hydrogen bonds
  • Weak hydrogen bonds
  • Resolved disulfide/covalent bonds and inferred potential contacts
• Calculate SASA (Solvent Accessible Surface Area) at atom, residue, and chain levels
• Calculate relative SASA (RSA) normalized by MaxASA values
• Calculate SAP (Spatial Aggregation Propensity) scores for aggregation prediction
• Calculate Shape Complementarity (SC) scores at protein-protein interfaces
• Superpose structures with Kabsch RMSD and cluster conformations with k-medoids
• Filter calculations to specific chains
• Output results in CSV, Parquet, or NDJSON
• Python bindings via PyO3
• Returns Polars DataFrames for efficient data manipulation

Scientific conventions

• Contact rows use Disulfide for resolved PDB SSBOND or mmCIF disulfide declarations and Covalent for other resolved LINK, CONECT, or _struct_conn bonds. Undeclared CYS pairs matching the original distance and CB--SG--SG--CB dihedral rule produce PotentialDisulfide; other contacts in the covalent-distance band produce PotentialCovalent. Clash and van der Waals regions are separately named.
• Explicit hydrogen-bond geometry uses only hydrogens associated with the donor atom. Missing donor hydrogens produce warnings; Arpeggia does not protonate input structures.
• Histidines use AllCharged by default for Arpeggio-compatible positive-ionisable typing. Heuristic applies explicit evidence followed by a pH-dependent intrinsic-pKa prior, while ExplicitOnly never guesses. Inferred histidine charge produces potential ionic, repulsion, and cation-pi labels rather than definitive ones.
• All analyses deterministically choose the highest-occupancy alternate conformer, with A as the tie-breaker, and warn when selection occurs.
• Standard atom, residue, and chain SASA use one atom population and ProtOr radii with elemental fallback. Polar/hydrophobic columns follow Rosetta's legacy SasaFilter atom partition; numerical areas remain Shrake–Rupley.
• dSASA is the two-sided buried area SASA(group 1) + SASA(group 2) - SASA(complex). Divide by two only when a one-sided interface-area convention is required.
• SAP uses the Rosetta-compatible full-atom Reduce-radius exposure definition with a 1.1 Å default probe and sums positive score contributions while reporting complete side-chain SASA. Arpeggia does not add missing atoms, so direct Rosetta comparison requires the same caller-prepared full-atom input. Monomers without a Rosetta calibration are omitted with a warning.
• RMSD uses uniform-weight Kabsch superposition with proper rotations and exact selected-atom correspondence. Structure clustering uses the resulting pairwise RMSD matrix and observed medoid structures; it does not perform sequence alignment or add missing atoms.

Installation

Python Package (Recommended)

Install using pip:

pip install arpeggia

Published wheels support CPython 3.10–3.14 on x86-64 Linux and Windows, and Arm64 Linux and macOS.

Or install from source using maturin:

git clone https://github.com/y1zhou/arpeggia.git
cd arpeggia
uv sync --frozen --all-extras
uv run maturin develop --uv --release --features python --locked

Rust Binary

For the command-line tool, you can install pre-built binaries from the GitHub Releases page, or build from source:

git clone https://github.com/y1zhou/arpeggia.git
cd arpeggia
cargo install --path .

This will install the arpeggia binary to your Cargo binary directory (usually ~/.cargo/bin).

Usage

Python API

import arpeggia

# Analyze protein contacts
contacts_df = arpeggia.contacts(
    "structure.pdb",
    groups="/",                    # All-to-all chain interactions
    vdw_comp=0.1,                 # VdW radii compensation
    dist_cutoff=6.5,              # Distance cutoff in Ångströms
    ignore_zero_occupancy=False,  # Set True to ignore zero occupancy atoms
    protonation="all-charged",    # Or "heuristic" / "explicit-only"
    ph=7.4,
)
print(f"Found {len(contacts_df)} contacts")
print(contacts_df.head())

# Calculate solvent accessible surface area
# Atom-level (default)
sasa_df = arpeggia.sasa("structure.pdb", level="atom", probe_radius=1.4, n_points=100, model_num=0)
print(f"Calculated SASA for {len(sasa_df)} atoms")

# Residue-level SASA
residue_sasa = arpeggia.sasa("structure.pdb", level="residue")
print(f"Calculated SASA for {len(residue_sasa)} residues")

# Chain-level SASA for specific chains only
chain_sasa = arpeggia.sasa("structure.pdb", level="chain", chains="A,B")
print(f"Calculated SASA for chains A and B")

# Calculate relative SASA (RSA) normalized by Tien et al. (2013) MaxASA values
rsa_df = arpeggia.relative_sasa("structure.pdb")
print(f"Calculated RSA for {len(rsa_df)} residues")

# Calculate Spatial Aggregation Propensity (SAP) scores for aggregation prediction
sap_df = arpeggia.sap_score("antibody.pdb", level="residue")
print(f"Calculated SAP for {len(sap_df)} residues")

# SAP for specific chains (e.g., antibody heavy and light chains)
sap_hl = arpeggia.sap_score("antibody.pdb", chains="H,L", sap_radius=5.0)
print(f"Calculated SAP for H and L chains")

# Calculate buried surface area at the interface
bsa = arpeggia.dsasa("structure.pdb", groups="A,B/C,D")
print(f"Buried surface area: {bsa:.2f} Ų")

# Additive two-sided polarity components
total, polar, hydrophobic, unknown = arpeggia.dsasa_components(
    "structure.pdb", groups="A,B/C,D"
)

# Calculate Shape Complementarity at an interface
sc_score = arpeggia.sc("antibody_antigen.pdb", groups="H,L/A")
print(f"Shape Complementarity: {sc_score:.3f}")  # Typical values: 0.5-0.7

# Extract protein sequences
sequences = arpeggia.seq("structure.pdb")
for chain_id, seq in sequences:
    print(f"Chain {chain_id}: {seq}")

# Extract declared SEQRES/entity-polymer sequences, including missing coordinates
declared_sequences = arpeggia.seqres("structure.pdb")

# Superpose on chain A, measure chains B/C, and cluster a pairwise matrix
rmsd = arpeggia.rmsd(
    "reference.cif",
    "mobile.cif",
    superpose_residues="A",
    rmsd_residues="B,C",
    atoms="ca",
)
pairs = arpeggia.pairwise_rmsd("structures/", num_threads=8)
clusters = arpeggia.cluster_structs(pairwise_rmsd=pairs, max_clusters=10)

The functions return Polars DataFrames for efficient data manipulation. You can easily convert to pandas if needed:

import polars as pl

# Convert to pandas
contacts_pd = contacts_df.to_pandas()

# Or save directly to various formats
contacts_df.write_csv("contacts.csv")
contacts_df.write_parquet("contacts.parquet")

Command-Line Interface

The CLI provides the same functionality:

# Analyze contacts
arpeggia contacts -i structure.pdb -o output_dir -g "A,B/C,D" -t csv

# Analyze contacts, ignoring atoms with zero occupancy
arpeggia contacts -i structure.pdb -o output_dir --ignore-zero-occupancy

# Calculate SASA at different levels (atom, residue, chain)
arpeggia sasa -i structure.pdb -o output_dir --level atom
arpeggia sasa -i structure.pdb -o output_dir --level residue
arpeggia sasa -i structure.pdb -o output_dir --level chain

# Calculate SASA for specific chains only
arpeggia sasa -i structure.pdb -o output_dir --level residue --chains "A,B"

# Calculate relative SASA (RSA) for each residue
arpeggia relative-sasa -i structure.pdb -o output_dir

# Calculate SAP scores for aggregation prediction
arpeggia sap -i antibody.pdb -o output_dir --level residue

# Calculate SAP for specific chains (e.g., antibody H and L chains)
arpeggia sap -i antibody.pdb -o output_dir --chains "H,L"

# Calculate buried surface area at the interface
arpeggia dsasa -i structure.pdb -g "A,B/C,D"

# Calculate Shape Complementarity at an interface
arpeggia sc -i antibody_antigen.pdb -g "H,L/A"

# Extract sequences
arpeggia seq structure.pdb

# Extract declared SEQRES/entity-polymer sequences
arpeggia seqres structure.pdb

# RMSD and fixed-count clustering; save the reusable long pair table
arpeggia rmsd reference.cif mobile.cif --atoms backbone
arpeggia cluster-structs -i structures/ -o results/ \
  --num-clusters 5 --pairwise-rmsd --num-threads 8

To see all available options:

arpeggia help
arpeggia contacts --help

See Structure RMSD and clustering for selection grammar, exact-correspondence requirements, output schemas, cache semantics, memory estimates, and threading behavior.

Chain Groups Specification

The groups parameter allows you to specify which chains interact with each other:

• "/" - All chains interact with all chains (including self) for contacts
• "A,B/C,D" - Chains A,B interact with chains C,D
• "A/" - Chain A interacts with all other chains
• "A,B/" - Chains A,B interact with all remaining chains

dSASA and SC require two disjoint, non-empty groups; "/" is therefore invalid for those calculations.

Development

To build the Python package in development mode:

uv sync --frozen --all-extras
uv run maturin develop --uv --features python --locked
uv run pytest

To run Rust tests:

cargo test --locked

License

GNU General Public License v3.0 - see LICENSE for details.

Credit

This project would not be possible without the following resources:

• Arpeggio: Original Python library for protein-protein interaction analysis.
• pdbtbx: The structural file parser doing all the heavy lifting.
• RustSASA: Library for calculating solvent accessible surface area.
• sc-rs: Library for calculating the Shape Complementarity by Lawrence & Colman (1993).
• Rosetta: Where the Spatial Aggregation Propensity (SAP) score calculations are inspired from.