Lipid Membranes & Protein-Loaded NPs (examples/aunps)

This tutorial demonstrates the Resolution-Agnostic capability of DoMD. We will reconstruct a highly complex system typical of Martini-level coarse-graining: a lipid membrane interacting with PEGylated Gold Nanoparticles (AuNPs) and protein-loaded Lipid Nanoparticles (LNPs).

Key concepts introduced here:

1. Martini-to-All-Atom Mapping: Mapping coarse-grained beads (representing chemical fragments like Phosphates, Glycerol, Alkyl tails) back to atomic resolution.

2. “Artificial” Reaction Templates: Using dummy atoms (Fluorine) as positioning markers to define connectivity between fragments.

3. Complex Heterogeneity: Simultaneously handling metals, polymers, lipids, and peptides in a single workflow.

Important

The “Artificial Chemistry” Technique

In this tutorial, the chemical reactions defined in reaction_template are NOT physically real reactions. They are topological instructions used to stitch fragments together.

Notice the frequent use of Fluorine (F) atoms in the SMILES strings (e.g., PEO = 'CCOF'). Here, ‘F’ acts as a “Connector Port” or “Positioning Marker”.

• Logic: The SMARTS pattern [C:1][F:2] + [C:3] >> [C:1][C:3] tells DoMD to bond the carbons and delete the ‘F’ atom.

• Purpose: This ensures that when fragments are assembled, they strictly maintain the correct orientation and valency, effectively “snapping” together like Lego blocks.

Step 1: Fragment Definitions (The Building Blocks)

We define the SMILES for every component. Note that these are not just whole molecules, but fragments corresponding to CG beads (e.g., Martini beads).

• Lipid Parts: Headgroups (NC3, PO4), Glycerol (GL1, GL2), Tails (C1A, D2A…).

• Peptide Residues: Specific amino acid fragments (LYS, TRP, VAL…).

• Nanoparticle: Gold (Au) and PEGylating agents (PEO).

from domd_tools import create_cg_system, build_aa_topology, assign_ff_parameters, get_gmx
from misc.logger import logger
logger.setLevel('INFO')

# --- 1. Define Components (SMILES) ---
# PEO chain and Au
PEO = 'OCCO'
Au = '[Au]'

# Peptide Residues (Targeting a specific sequence)
LYS1 = 'NCCCC[C@@H](N)C(=O)O'
VAL1 = 'CC(C)[C@@H](N)C(=O)O'
TRP1 = 'c1cccc2c(C[C@@H](N)C(=O)O)cnc2c1'
# ... (Refer to the full script for all residues) ...

# Lipid Fragments (Martini mapping logic)
NC3 = 'C[N+](C)(C)C'        # Choline head
PO4 = 'OP(=O)([O-])OC'      # Phosphate
GL1 = 'FC[C@@H](OC(=O)CC)CO' # Glycerol backbone (with F marker)
GL2 = 'C(=O)CC'
GL0 = 'OCCO'
C1A = 'CCCF'                # Saturated tail fragment
D2A = 'CCC=CCC'             # Unsaturated tail fragment
# ... (Refer to the full script for all lipid parts) ...

Step 2: All-Atom Connectivity Definitions (The Reaction Template)

This is the core of the resolution-agnostic approach. We define how these fragments connect to form the full molecules (Lipids, Peptides, PEO-Au).

Observe the SMARTS pattern: [CH3:1].[C:2][F:3]>>[CH2:1][C:2].[F:3]. This translates to: “Take a Methyl group, connect it to the Carbon holding the Fluorine, and eject the Fluorine.”

# --- 2. Define Connectivity Rules ---
reaction_template = {
    # --- Polymerization & Grafting ---
    'a': { # PEO-PEO chain growth
        'cg_reactant_list': [('PEO', 'PEO')],
        'smarts': '[C:1][O:2].[O:3][C:4]>>[C:1][O:2][C:4].[O:3]',
        'prod_idx': [0]
    },
    'b': { # Grafting PEO to Gold Surface
        'cg_reactant_list': [('Au', 'PEO')],
        'smarts': '[Au:1].[OH1:2][C:3]>>[Au:1][O:2][C:3]',
        'prod_idx': [0]
    },

    # --- Lipid Assembly (Stitching Head-Glycerol-Tails) ---
    'NC3-PO4': {
        'cg_reactant_list': [('NC3', 'PO4')],
        'smarts': '[C:1].[C:2]>>[C:1][C:2]',
        'prod_idx': [0]
    },
    'PO4-GL1': {
        'cg_reactant_list': [('PO4', 'GL1')],
        'smarts': '[O:1].[C@@H:2][C:3][F:4]>>[O:1][C:3][C@@H:2].[F:4]',
        'prod_idx': [0]
    },
    'GL1-C1A': { # Connecting Tail to Glycerol
        'cg_reactant_list': [('GL1', 'C1A')],
        'smarts': '[CH3:1].[C:2][F:3]>>[CH2:1][C:2].[F:3]',
        'prod_idx': [0]
    },

    # --- Peptide Synthesis (Peptide Bonds) ---
    'VAL1-TRP1': {
        'cg_reactant_list': [('VAL1', 'TRP1')],
        'smarts': '[NH2:1][C@@H1:2].[C:3](=[O:4])[OH1:5]>>[C:3](=[O:4])[NH1:1][C@@H1:2].[OH2:5]',
        'prod_idx': [0]
    },
    # ... (Additional peptide bonds for LYS, TRP, etc.) ...
}

Step 3: All-Atom Topology Building

We load the fragments into the mols dictionary. Then, we perform backmapping on a pre-existing CG snapshot (cg/lipidmem_with_AuNP.xml). This snapshot represents the equilibrated state of the Martini-level simulation.

# --- 3. Register Molecules ---
# (Full dictionary from the script)
mols = {
    'PEO': {'smiles': PEO, 'file': None, 'is_rigid':False},
    'Au': {'smiles': Au, 'file': None, 'is_rigid':False},
    'LYS1': {'smiles': LYS1, 'file': None, 'is_rigid':False},
    # ... (Full list of molecules) ...
    'C1A': {'smiles': C1A, 'file': None, 'is_rigid':False},
    'W': {'smiles': 'O', 'file': None, 'is_rigid':False},
}

# --- 4. Backmapping ---
# Input: A Martini-scale CG configuration
xmlfile = 'cg/lipidmem_with_AuNP.xml'

# Reconstruct AA Topology
# large=50 ensures spatial decomposition is used for large membrane/NP structures
confs, mol_graphs, box = build_aa_topology(
    mols,
    reaction_template,
    xmlfile,
    large=50
)

Figure 1: Visualization of the backmapping process. DoMD replaces the coarse-grained beads (left) with the defined chemical fragments (right), removing the ‘F’ connectors to form continuous molecules.

Step 4: ML Force Field & Output

Given the complexity and heterogeneity of the system (Lipids + Gold + Peptides), we employ the Machine Learning (ML) force field strategy. This ensures that parameters for the specific lipid fragments and the gold interface are handled consistently without manual atom-typing.

# --- 5. Force Field Assignment ---
# Use Pure ML strategy for all components except single atoms (e.g., Au, ions) which can use TPL
strategies = []
for mol_graph in mol_graphs:
    if mol_graph.number_of_nodes() == 1:
        strategies.append('tpl')
    else:
        strategies.append('ml')

ffs = assign_ff_parameters(mol_graphs, strategies=strategies)

# --- 6. Export ---
get_gmx(mol_graphs, confs, ffs, box)

Final Result: DoMD successfully translates the “beads” of the Martini-like input into fully connected, chemically accurate all-atom structures of:

1. A Phospholipid Bilayer.

2. Gold Nanoparticles with grafted PEG chains.

3. Peptide/Protein chains embedded or interacting with the system.