Skip to main content

NP-Likeness Scoring

NP-likeness (Natural Product likeness) scoring evaluates whether a molecule resembles known natural products or synthetic compounds.

What It Measures​

The NP-likeness score is based on structural fragments and their frequency in:

  • Natural product databases (COCONUT, Dictionary of Natural Products)
  • Synthetic molecule databases (PubChem, ZINC)

Higher scores indicate natural product-like character, lower scores indicate synthetic-like character.

Score Range and Interpretation​

ScoreCategoryDescription
≥ 2.0Strong NP-likeNatural product features clearly evident
1.0–2.0NP-likeSuggestive of natural origin
0.3–1.0Moderate NP-likeSome NP features present
−0.3 to 0.3MixedBoth NP and synthetic characteristics
−1.0 to −0.3Moderate syntheticMore synthetic than NP features
−2.0 to −1.0Synthetic-likeTypical synthetic compound profile
< −2.0Strong syntheticLacks natural product features

Color scale in the UI: green (NP-like) → slate (mixed) → red (synthetic).

Typical Scores​

Natural products:

  • Morphine: +2.5
  • Taxol: +3.8
  • Quinine: +2.1

Synthetic drugs:

  • Ibuprofen: -1.8
  • Aspirin: -0.9
  • Lipitor: -2.3

API Usage​

curl -X POST http://localhost:8001/api/v1/score \
  -H "Content-Type: application/json" \
  -d '{
    "molecule": "CN1CCC23C4C1CC5=C2C(=C(C=C5)O)OC3C(C=C4)O",
    "include": ["np_likeness"]
  }'

Response:

{
  "np_likeness": {
    "score": 2.1,
    "interpretation": "Natural product-like molecule",
    "caveats": []
  }
}

Caveats and Limitations​

Molecule size effects:

For very large molecules (>50 heavy atoms), scores may be less reliable. The model is trained on typical drug-sized molecules.

Fragment-based:

The score is based on structural fragments. Molecules with unusual fragment combinations may have unexpected scores.

Training data bias:

Scores reflect the training data composition. Rare natural product classes may score as synthetic if underrepresented in training data.

Caveat Reporting

When limitations apply, they're reported in the caveats array for transparency.

Use Cases​

Natural Product Research​

Identify natural product-like molecules in screening libraries:

np_likeness_score > 1.0

Diversity Analysis​

Assess chemical diversity by NP-likeness distribution:

  • Wide range (-3 to +3): High diversity
  • Narrow range: Homogeneous library
  • Skewed distribution: Biased toward NP or synthetic space

Scaffold Analysis​

Combine with scaffold analysis to identify natural product-inspired scaffolds:

np_likeness_score > 1.0 AND
has_scaffold = true

Library Design​

Balance natural product and synthetic character:

  • NP-focused library: Score > 0
  • Synthetic-focused library: Score < 0
  • Balanced library: Score -1 to +1

Interpretation Guidelines​

High positive score (> 2.0):

  • Likely contains natural product fragments
  • May have complex ring systems
  • Often higher molecular weight
  • Potentially harder to synthesize

Near zero (-1.0 to +1.0):

  • Intermediate character
  • May be NP-inspired synthetic molecules
  • Simple aromatic compounds
  • Modified natural products

High negative score (< -2.0):

  • Synthetic-like fragments dominant
  • Simpler structure
  • Drug-like properties common
  • Easier to synthesize

Relationship to Other Scores​

NP-likeness vs Drug-likeness:

  • Natural products often violate Lipinski
  • Higher complexity (more rings, stereocenters)
  • Often larger molecular weight

NP-likeness vs Synthetic Accessibility:

  • Positive correlation: NP-like → harder to synthesize
  • Not absolute: some NPs are easy to synthesize

NP-likeness vs Scaffold Complexity:

  • NP-like molecules often have complex scaffolds
  • Multiple fused ring systems
  • More stereocenters

Best Practices​

  1. Use as a guide: Not a definitive natural product classifier
  2. Combine with other data: Check COCONUT database for confirmation
  3. Consider therapeutic area: Some drug classes are NP-like (antibiotics)
  4. Track during optimization: Monitor changes in NP character
  5. Balance with drug-likeness: High NP-likeness may sacrifice drug-likeness

Reference​

Ertl, P., Roggo, S. & Schuffenhauer, A. (2008). Natural product-likeness score and its application for prioritization of compound libraries. Journal of Chemical Information and Modeling, 48(1), 68–74.

Next Steps​