Field notes
Methodology essays, architecture decisions, validation deep-dives, and the regulatory context behind the design choices. Curated by the MolTrace team — written for analysts, engineers, and regulatory reviewers who want the actual reasoning.
Featured essay
MethodologyThe expert-reference vs detector-output mismatch that kept our promotion gate red — and the multiplet-clustering layer that reconciled it.
Δ=17 → Δ=2
Illustrative — sample figures
Median absolute peak-count delta against the NMRShiftDB2 reference, set against the ≤2 the strict gate asks for.
Editorial streams
We publish across science, engineering, and methodology — each stream has its own audience but shares the same rigor.
Science
Methodology essays, validation deep-dives, and notes from the analytical team.
Engineering
Architecture decisions, contract design, perf wins, and the instrumentation under the hood.
Methodology
How we measure ourselves. Promotion gates, regression-corpus design, and what 'experimental' really means.
How a 20-fixture A/B JSON sidecar replaced our 'looks-good-to-me' detector reviews.
Every detector change runs against a curated NMRShiftDB2 corpus before merge. CI fails by name when any single fixture drifts >50% — so reviewers see 'nmrshiftdb2_60000006_13c regressed' instead of 'tests passed (with notes).' The boring infrastructure meant to keep ship velocity high.
A new analysis backend ships opt-in behind two published numbers. Promotion happens when the numbers move, and the commit that removes the failure marker is the record.
Our GSD sidecar shipped as experimental: true with a written promotion gate — 95% solvent auto-detection and a median compound-environment-count delta of 2 or better. The gate was declared before it was met, the test that enforced it was marked expected-to-fail, and clearing it is a diff you can read.
Why we'd rather show 'pending' than a polished score with no provenance.
Every numerical claim in the UI links to its source — the spectrum file, the picked peaks, the SMILES candidate, the literature citation, the human reviewer who signed off. The implementation cost is real. The regulatory cost of doing it otherwise is higher.
A 500-MHz field hardcoded in the FE became a real number from the vendor metadata. Three lines of code, one cascade, no contract change.
Phase 8 traced field_mhz through the preview → process → analyze chain so the GSD endpoint receives the spectrometer frequency the instrument actually used (600.13 MHz, in our verification fixture) instead of a hardcoded 500. The same plumbing pattern works for vendor / solvent / nucleus.
If a reference table and a detector count in different units, no threshold you pick will mean anything. So we built a corpus that counts both ways.
A published peak list and a deconvolution engine disagree by construction. The HMDB-style harness forward-models a spectrum from a reference list and then gates on environment-count and multiplet-line-count deltas separately, so detector quality can be separated from corpus granularity. It is also why one of our three corpora is deliberately not gated on peak count at all.
We publish on a deliberate cadence — methodology essays land on shipping milestones, not on a content calendar. No marketing emails, no upsells. Just the writing.
No tracking pixels · Unsubscribe in one click · Designed to support GDPR-aligned data handling