Compound Data Format Conversion Tool (SMILES, One-line Graph, gSpan Format)

This tool is an integrated CLI for converting compound data (SMILES format, one-line graph CSV, gSpan format) between different formats. The converted data can be directly used in pmopt and other tools.

---

Required Python Packages

• pysmiles: SMILES notation molecular reading
• pandas: DataFrame operations
• networkx: Graph operations
• rdkit: Molecular structure conversion

Installation example:

pip install pysmiles pandas networkx rdkit

---

Usage (converter_cli.py)

Basic Commands

python converter_cli.py --mode smiles_to_graph --input input.csv --output output.csv --smiles-column SMILES --objective-column y --hydrogen 0
python converter_cli.py --mode graph_to_gspan --input input.csv --output output.graph
python converter_cli.py --mode gspan_to_graph --input input.graph --output output.csv
python converter_cli.py --mode graph_to_smiles --input input.csv --output output.csv

Mode Descriptions

1. SMILES → One-line Graph CSV

• --mode smiles_to_graph
• Required: --input input CSV/Excel, --output output CSV
• Optional: --smiles-column SMILES column name, --objective-column objective variable column name, --hydrogen include hydrogen (1/0)

2. One-line Graph CSV → gSpan Format

• --mode graph_to_gspan
• Required: --input input CSV, --output output .graph
• Optional: --graph-column graph column name, --spacing blank lines between graphs, --target-column objective variable column name (output in header)

3. gSpan Format → One-line Graph CSV

• --mode gspan_to_graph
• Required: --input input .graph, --output output CSV
• Optional: --gspan-target-column objective variable column name (default: objective)

4. One-line Graph CSV → SMILES

• --mode graph_to_smiles
• Required: --input input CSV, --output output CSV
• Optional: --graph-column graph column name, --other-columns other column names to output together

---

Input/Output Examples

SMILES→One-line Graph CSV

Input CSV:

objective,SMILES
-5.89,O=C1c3ccccc3[Se]N1c2ccccc2

Command:

python converter_cli.py --mode smiles_to_graph --input input.csv --output output.csv --smiles-column SMILES --objective-column objective --hydrogen 1

Output CSV:

objective,graph
-5.89,v 0 O v 1 C ... e 0 1 = ...

One-line Graph CSV→gSpan

Input CSV:

graph,objective
v 0 O v 1 C ... e 0 1 = ...,100

Command:

python converter_cli.py --mode graph_to_gspan --input input.csv --output output.graph --target-column objective

Output .graph:

t # 0 objective 100
v 0 8
v 1 6
...
e 0 1 4
...

gSpan→One-line Graph CSV

Input .graph:

t # 0 objective 100
v 0 8
v 1 6
...
e 0 1 4
...

Command:

python converter_cli.py --mode gspan_to_graph --input input.graph --output output.csv

Output CSV:

objective,graph
100,v 0 O v 1 C ... e 0 1 = ...

One-line Graph CSV→SMILES

Input CSV:

graph
v 0 O v 1 C ... e 0 1 = ...

Command:

python converter_cli.py --mode graph_to_smiles --input input.csv --output output.csv

Output CSV:

smiles
O=C1c3ccccc3[Se]N1c2ccccc2

---

Option Details

• --smiles-column : Column name containing SMILES (default: smiles)
• --objective-column : Objective variable column name (default: objective)
• --hydrogen : Whether to include hydrogen (1: include, 0: exclude, default: 0)
• --graph-column : Column name containing one-line graph (default: graph)
• --spacing : Insert blank lines between graphs in gSpan output
• --target-column : Column name for objective variable (target value) to output in gSpan header (e.g., objective, y, etc., optional)
• --gspan-target-column : Objective variable column name for gSpan input (default: objective)
• --other-columns : Other column names to output together when generating SMILES

---

Important Notes

• Please pay attention to the column names and format of input files
• Errors will occur if there are unknown element symbols or bond symbols
• If there are multiple input SMILES, the longest one will be selected
• Converted data can be directly used in pmopt

【Important】Information Loss and Limitations in Reverse Conversion

• When converting SMILES→One-line Graph, the following information is lost:
  • Stereochemical information ([C@H], E/Z, etc.)
  • Explicit aromaticity flags
  • Number of hydrogens (when hydrogen=0)
  • Atomic charge and radical information
• Therefore, reverse conversion from One-line Graph→SMILES may not completely match the original SMILES.
• Particularly for molecules containing aromatic rings or stereochemistry, RDKit's Kekulize processing may cause errors (Can't kekulize mol.), making reverse conversion to SMILES impossible.
• This is due to the specifications of this tool and RDKit/pysmiles, and depends on the structure and information content of the input molecule.
• If complete reverse conversion is required, it is recommended to also save the original SMILES.

---

SMILES to SVG Image Generation (draw_smiles_to_svg.py)

A tool for visualizing molecular structures in SMILES notation as SVG images.

Basic Commands

python draw_smiles_to_svg.py input.csv output_dir
python draw_smiles_to_svg.py input.csv output_dir --smiles_col SMILES

Argument Descriptions

• csv_path: Input CSV file (file containing SMILES)
• out_dir: Output directory (where SVG files will be saved)
• --smiles_col: Column name containing SMILES (default: smiles)

Usage Example

Input CSV (input.csv):

smiles,name
O=C1c3ccccc3[Se]N1c2ccccc2,compound1
CC(C)CC1=CC=C(C=C1)C(C)C(=O)O,compound2

Command:

python draw_smiles_to_svg.py input.csv svg_output

Output:

• svg_output/0.svg: SVG image of molecular structure from row 1
• svg_output/1.svg: SVG image of molecular structure from row 2

Important Notes

• If there are invalid SMILES notations, those rows will be skipped and error messages will be displayed
• Output SVG images are 300x300 pixels in size
• If the output directory doesn't exist, it will be created automatically