nmr-analyze-simulate

NMR analysis and simulation

Preserve the distinction between observations, assignments, calculations, and hypotheses. Never present synthetic data as experimental data.

Route the request

1. Identify the input:
   • For a screenshot or plotted spectrum, perform visual/semiquantitative interpretation only.
   • For CSV/JCAMP/peak tables, analyze the supplied processed values and retain their units.
   • For a directory containing fid + procpar, treat it as Varian/Agilent.
   • For a directory containing fid/ser + acqus, treat it as Bruker.
   • For a molecular structure without measured data, predict plausible regions and label every assignment as predicted.
2. Choose the task:
   • For interpretation or raw-data processing, follow Interpret a spectrum.
   • For a ChemDraw structure plus measured 1D NMR that must be labeled and checked in MestReNova, follow Assign and verify in MestReNova.
   • For synthetic data, follow Simulate a raw FID.
   • For comparison, process measured and synthetic data independently, then compare peak positions, multiplicities, relative integrals, and missing/extra signals.
3. Read references/interpretation.md for assignment, quantitation, uncertainty, and reporting rules.
4. Read references/mnova-structure-verification.md before writing assignments into MestReNova.
5. Read references/simulation.md before constructing or reviewing a peak-model JSON.

Interpret a spectrum

1. Work on a copy or write all outputs to a new directory. Do not modify vendor raw data.
2. Record nucleus, solvent, field strength, temperature, pulse program, scans, acquisition time, relaxation delay, spectral width, and reference when available. Mark missing metadata.
3. Process 1D raw FIDs with scripts/process_1d.py. Before installing anything, search the current project and the input dataset's ancestors for .venv, venv, or .nmr_env and use its Python when it contains numpy, scipy, matplotlib, and nmrglue. Otherwise use the bundled workspace Python or system Python. Do not install dependencies into the analysis-output directory.
4. Inspect the full spectrum before zooming into diagnostic regions. Check phase, baseline, truncation/ringing, solvent suppression, clipping, digital-filter artifacts, water/solvent peaks, and signal-to-noise.
5. Reference against a stated internal standard, residual solvent, or known signal. Do not silently assume a reference.
6. Build a peak table with chemical shift/range, multiplicity, J values when resolved, integral, tentative assignment, evidence, alternatives, and confidence.
7. Test assignments against all constraints: proton/carbon count, symmetry, multiplicity, coupling, integration, expected shift, exchange, overlap, isotope labeling, and the proposed structure or reaction.
8. Separate conclusions into:
   • directly observed;
   • strongly supported assignment;
   • tentative/ambiguous assignment;
   • not assessable from the supplied data.
9. For quantitative claims, show the normalization signal and formula. Check relaxation delay, saturation, overlap, baseline, and response assumptions before reporting conversion, composition, DP, or purity.
10. Recommend the smallest decisive follow-up experiment, such as a longer relaxation delay, more scans, COSY, HSQC, HMBC, DEPT/APT, DOSY, variable temperature, spike-in, or an orthogonal method.