SpectraCheck Overview
SpectraCheck turns raw analytical evidence into reviewed interpretations. It supports NMR, LC-MS/MS, and quantitative workflows while preserving the original file archive for audit.
- 1H NMR, 13C NMR, and 2D NMR workflows
- LC-MS/MS annotation and compound-family grouping
- Raw FID archive preservation
- Peak classification as compound, solvent, impurity, or reference
- Confidence scores, contradiction flags, and report-ready evidence
Authoring plan
Section titled “Authoring plan”Before this guide is marked production-ready, interview the NMR scientist for one hour. Ask:
- What are the five most common tasks users perform in SpectraCheck?
- What are the three most common user mistakes?
- Which instrument metadata fields are most often missing or wrong?
- Which results are safe to accept quickly, and which always need a deeper review?
Use the interview to write the workflow in the scientist’s words, then have the same owner review the final page.
Supported techniques
Section titled “Supported techniques”| Technique | SpectraCheck support |
|---|---|
| 1H NMR | Peak classification, integration review, impurity detection, and solvent/reference checks. |
| 13C NMR | Carbon assignment, solvent identification, and structure-support evidence. |
| 2D NMR: COSY, HSQC, HMBC | Connectivity mapping and structure elucidation support. |
| LC-MS/MS | Fragmentation annotation, unknown identification, and compound-family grouping. |
| qNMR | Absolute quantification workflows with USP <761> context after expert review. |
Supported source files
Section titled “Supported source files”| Instrument or format | Required files |
|---|---|
| Bruker TopSpin 1D | fid plus acqus. |
| Bruker TopSpin 2D | ser plus acqu2s and related acquisition metadata. |
| Agilent | Raw fid binary and available acquisition metadata. |
| JCAMP-DX | .jdx file when raw FID evidence is unavailable. |
Always upload the parameter file with raw FID data. Missing parameters reduce interpretability and should create a visible warning.
Processing pipeline
Section titled “Processing pipeline”- Read the immutable source archive.
- Process a derived copy of the data.
- Detect peaks and relevant regions.
- Classify peaks as compound, solvent, impurity, reference, or unknown.
- Score the interpretation and surface contradictions for human review.
Recent capabilities
Section titled “Recent capabilities”The SpectraCheck analysis backend has expanded significantly. The shipped surface today includes:
- Opt-in Global Spectral Deconvolution with auto-classification (
POST /spectrum/analyze/gsd), promotion-gate-cleared on NMRShiftDB2 and validated against a 100-fixture real-instrument HMDB corpus (93 % solvent auto-detect). - Multiplet analysis (
POST /spectrum/analyze/multiplets) with J-coupling recovery and synthetic predicted-vs-observed overlay. - Chemical-shift prediction (
POST /spectrum/predict/shifts) with an NMRNet SE(3)-equivariant backend and a HOSE-code topological fallback. - Automated structure verification (ASV) combining four independent tests into one auditable posterior confidence.
- Quantitative integration with Sum / Edited Sum / Peaks (
POST /spectrum/analyze/integration) — Edited Sum recovers the true compound integral within 1 % at impurity fractions up to 25 %. - Spectrum retrieval (
POST /spectrum/retrieve) with FAISS HNSW vector + Kuhn-Munkres set similarity (top-100 from 45 k in ≈ 2 ms).
See NMR Interpretation for the full capability catalog with validation numbers, and qNMR Quantification for the integration methods.