SMILES Parsing

The first step of the pipeline converts a SMILES (Simplified Molecular Input Line Entry System) string into a molecular graph suitable for 3D coordinate generation.

From String to Graph

The Molecule Structure

The parser produces a Molecule containing:

pub struct Molecule {
    pub graph: Graph<Atom, Bond, Undirected>,
    pub ring_info: Vec<Vec<usize>>,  // SSSR ring membership
}

Atom Properties

Each atom stores the properties needed for 3D generation:

Property	Type	Description
element	u8	Atomic number (1=H, 6=C, 7=N, 8=O, ...)
hybridization	Hybridization	SP, SP2, SP3, or Unspecified
formal_charge	i8	Formal charge (−2 to +2 typically)
aromatic	bool	Whether the atom is aromatic
num_implicit_h	u8	Implicit hydrogens added
chiral	Option<Chirality>	@ (CCW) or @@ (CW)
in_ring	bool	Whether the atom is in any ring

Bond Properties

Property	Type	Values
order	BondOrder	Single, Double, Triple, Aromatic
stereo	BondStereo	None, E (trans), Z (cis)

Hybridization Assignment

After building the graph, hybridization is assigned based on local environment:

$$\text{hybridization}=\{\begin{array}{ll}\text{SP3}& \text{if all bonds are single}\\ \text{SP2}& \text{if any bond is double or atom is aromatic}\\ \text{SP}& \text{if any bond is triple or two double bonds}\end{array}$$

Special cases:

• Aromatic atoms → SP2
• N with 3 single bonds → SP2 if in an aromatic ring, SP3 otherwise
• O with 2 single bonds → SP3 (e.g., ether)
• Terminal atoms (degree 1) → inherit from neighbor

Implicit Hydrogen Addition

SMILES encodes most hydrogens implicitly. The parser adds them explicitly because 3D coordinate generation requires all atoms:

$$n_H=v-\sum _{bonds}\text{order}-|q|$$

where $v$ is the normal valence, and $q$ is the formal charge.

Standard valences used:

Element	Valences
C	4
N	3, 5
O	2
S	2, 4, 6
P	3, 5
B	3
F, Cl, Br, I	1
Si	4
Se	2

For aromatic atoms, the aromatic bond contributes 1.5 to the bond order sum, but in practice the parser assigns 1 per aromatic bond and adjusts the hydrogen count to maintain proper valence.

Ring Detection (SSSR)

The Smallest Set of Smallest Rings (SSSR) is computed after graph construction. This is essential for:

• Bounds matrix: ring bonds constrain torsion angles
• SMARTS matching: ring-membership queries (R, r)
• ETKDG: ring torsion patterns differ from chain torsions
• Force field: ring planarity enforcement

The SSSR is found using a modified graph traversal:

1. Compute the cycle rank: $\mu =|E|-|V|+1$
2. Find all shortest-path back edges
3. Extract $\mu$ independent rings

Example: Phenol

For the SMILES c1ccccc1O:

Result: 13 atoms (6C + 1O + 6H), 13 bonds, 1 SSSR ring of size 6, all ring carbons SP2, oxygen SP3.