Configuration

MOLRAPTOR reads all settings from a single YAML file passed via --config. The bundled example is examples/example_config.yaml.

Example Configuration

paths:
  raw_input_file: data/dataset.csv
  raw_output_file: data/properties.csv
  curated_output_file: data/properties_curated.csv
  error_log_file: logs/error_cids.txt
  fingerprint_output_file: artifacts/morgan_fp.csv
  fingerprint_array_file: artifacts/morgan_db.npy
  labels_output_file: artifacts/labels.npy

pubchem:
  properties:
    - MolecularWeight
    - XLogP
    - HBondDonorCount
    - HBondAcceptorCount
    - RotatableBondCount
    - TPSA
    - Complexity
    - SMILES
  timeout: 5
  max_retries: 3
  sleep_seconds: 0.2
  chunk_size: 400

fingerprint:
  radius: 2
  size: 1024

curate:
  required_columns:
    - PubChem CID
    - Label
    - MolecularWeight
    - Complexity
    - SMILES
  dtype_map:
    Label: int64

Use the file from the CLI:

molraptor run --config examples/example_config.yaml

Input Layout

MOLRAPTOR expects a CSV file with at least a PubChem CID and Label column:

data/
└── dataset.csv      <- input file with PubChem CIDs and labels

Minimum required columns:

Column	Description
PubChem CID	PubChem compound identifier
Label	Binary class label (0 or 1)

Configuration Sections

paths

Key	Description
raw_input_file	Input CSV with CIDs and labels
raw_output_file	Merged CSV after PubChem fetch
curated_output_file	Filtered CSV after curation
error_log_file	Log of failed CIDs during fetch
fingerprint_output_file	Morgan fingerprints as CSV
fingerprint_array_file	Morgan fingerprints as NumPy array
labels_output_file	Target labels as NumPy array

pubchem

Key	Default	Description
properties	—	List of PubChem properties to fetch
timeout	5	HTTP request timeout in seconds
max_retries	3	Maximum retry attempts per request
sleep_seconds	0.2	Delay between chunk requests
chunk_size	400	Number of CIDs per API request

fingerprint

Key	Default	Description
radius	2	Morgan fingerprint radius
size	1024	Fingerprint bit vector size

curate

Key	Description
required_columns	Columns that must be present and non-null
dtype_map	Column type enforcement (e.g. Label: int64)

Outputs

MOLRAPTOR writes all artifacts to artifacts/:

artifacts/
├── morgan_fp.csv      # Morgan fingerprints (human-readable)
├── morgan_db_*.npy    # Morgan fingerprints (NumPy array, shape: N×size)
├── labels.npy         # Target labels (NumPy array, shape: N,)
└── summary.txt        # Execution report

Output	Purpose
morgan_fp.csv	Human-readable fingerprint matrix
morgan_db_*.npy	ML-ready fingerprint array for downstream models
labels.npy	Target label vector aligned with fingerprints
summary.txt	Pipeline execution report with dataset statistics