Troubleshooting

Invalid SMILES string

The SMILES notation could not be parsed.

Fix: Check for unmatched parentheses, invalid atom symbols, or incorrect ring closure numbers. Use a SMILES validator (e.g., RDKit) to identify the specific error. Common mistakes:

• Missing ring closure digit: c1ccccc → should be c1ccccc1
• Unmatched brackets: [NH2 → should be [NH2]
• Invalid element: Xx is not a valid element symbol

CSV missing "smiles" column

Batch upload requires a column named smiles.

Fix: Ensure your CSV header row contains a column named exactly smiles (case-insensitive). Other common column names like SMILES, canonical_smiles, or structure are also accepted.

SDF file parse error

The uploaded SDF file contains malformed records.

Fix: Check that each molecule block ends with M END and records are separated by $$$$. Ensure the file uses V2000 or V3000 format. Try re-exporting from your chemical drawing tool.

Unsupported element in material composition

The Materials module does not recognize one or more elements.

Fix: FluxMateria supports 73 elements (H through Bi). Lanthanides (La–Lu) and elements beyond Bi are not currently supported. Check for typos in element symbols.

Check which endpoints are low-confidence

Confidence is per-endpoint, not per-molecule. A molecule might have high-confidence solubility but low-confidence hERG. Use the reliable endpoints for screening and flag the others for experimental follow-up.

Don't use low-confidence results as pass/fail gates

Rejecting a candidate based on a low-confidence prediction may discard a viable compound. Instead, flag it for experimental validation.

Look for structural features that reduce coverage

Common causes of low confidence: macrocycles, metallocenes, boron-containing compounds, very large molecules (MW > 800), heavily fluorinated scaffolds, or unusual heterocyclic systems.

Report persistent issues

If you're seeing low confidence across an entire chemical series that you believe should be well-covered, contact us. This helps us identify physics coverage gaps.