Chem Nmr Analysis

NMR Mixture Analysis

When to Use This Skill

The agent should use this skill's scripts when:

• A workflow (e.g., reaction-to-nmr-quantification.md or nmr-reaction-kinetics.md) calls for deconvolution, product prediction, kinetics analysis, or spectral plotting.
• The user already has reference spectra and a mixture spectrum and wants to quantify component proportions directly.
• The user has multiple time-point spectra and wants to track reaction progress via NMR.

For end-to-end workflows that chain this skill with other skills, see: .agents/workflows/reaction-to-nmr-quantification.md and .agents/workflows/nmr-reaction-kinetics.md.

When NOT to Use This Skill

• 13C NMR, 2D NMR (COSY, HSQC, etc.), or solid-state NMR -- this skill handles 1H solution-state NMR only.
• Structure elucidation of unknown compounds -- this skill requires knowing (or predicting) what compounds are in the mixture. It does not identify unknowns from scratch.
• Pure compound characterization -- if the user has a single pure compound and just wants to assign peaks, this skill is not appropriate. The agent should interpret the spectrum directly.
• Mass spectrometry data -- despite the Wasserstein algorithm's origins in mass spec, this skill operates on NMR chemical shift axes only.
• Digitizing spectrum images -- the agent should use the general-plot-digitizer skill for that step.
• Predicting NMR spectra from SMILES -- the agent should use the chem-nmr-predict skill for that step.
• Resolving compound names to SMILES -- the agent should use the drug-db-pubchem skill for that step.

Scripts Reference

predict_products.py

The agent should use this script to predict reaction products from reactant and reagent SMILES via the ReactionT5 model.

export HF_TOKEN=<token>
${CLAUDE_SKILL_DIR}/../../venv/run cpu python ${CLAUDE_SKILL_DIR}/scripts/predict_products.py \
  --reactant_smiles "C1CCC(=O)C1" \
  --reagent_smiles "[BH3-]" \
  --output <research_dir>/predicted_products.json

deconvolve.py

The agent should use this script to determine mole fractions of known components in a mixture spectrum via Wasserstein-distance deconvolution.

${CLAUDE_SKILL_DIR}/../../venv/run cpu python ${CLAUDE_SKILL_DIR}/scripts/deconvolve.py \
  mixture.csv ref_borneol.xy ref_isoborneol.xy \
  --protons 18 18 \
  --names "borneol" "isoborneol" \
  --baseline-correct \
  --plot <research_dir>/deconvolution_result.png \
  --json

kinetics.py

The agent should use this script when the user has crude NMR spectra recorded at multiple time points during a reaction.

${CLAUDE_SKILL_DIR}/../../venv/run cpu python ${CLAUDE_SKILL_DIR}/scripts/kinetics.py \
  --refs ref1.xy ref2.xy \
  --timepoints t0.csv t10.csv t20.csv \
  --times 0 10 20 \
  --time_unit min \
  --protons 18 18 \
  --names "reactant" "product" \
  --baseline_correct \
  --output_dir <research_dir>/kinetics/

plot.py

The agent should use this script to overlay or stack spectra for visual inspection before or after deconvolution.

${CLAUDE_SKILL_DIR}/../../venv/run cpu python ${CLAUDE_SKILL_DIR}/scripts/plot.py \
  mixture.csv ref_borneol.xy ref_isoborneol.xy \
  --labels "Mixture" "borneol" "isoborneol" \
  --title "Mixture vs References" \
  --output <research_dir>/spectra_overview.png

Key Arguments for deconvolve.py

Interpreting Results

The deconvolution output contains proportions and a Wasserstein distance (WD) indicating fit quality.

If/Then rules for Wasserstein distance:

• If WD < 0.05 -- good fit. The agent should report proportions with confidence.
• If 0.05 < WD < 0.15 -- acceptable fit. The agent should report proportions but note the fit quality and suggest possible causes (minor missing components, baseline noise).
• If WD > 0.15 -- poor fit. The agent should:
  1. Check if a component is missing (compare overlay plot for unmatched peaks).
  2. Check if there is a ppm calibration offset between mixture and references.
  3. Ask the user if there are additional species in the mixture not accounted for.
  4. Not report proportions as reliable.

If proportions do not sum to ~1.0 -- the agent should note that the "noise" fraction represents unmatched signal and explain what it might be.

Verification: After deconvolution, the agent must inspect the deconvolution plot, check the residual panel for large residuals, and verify that proportions are chemically reasonable. If results contradict known chemistry, the agent should flag this to the user rather than silently accepting.