MolTrace turns every spectrum into transparent, reproducible evidence. Confirm structures, elucidate unknowns, run a busy core facility, and generate publication-ready supporting information — with the reasoning kept visible and the trail back to raw data intact.
Why academic labs
Academic labs don't have a compliance department to absorb the busywork. They need reproducibility, teaching transparency, facility-wide consistency, and budget discipline — from the same tool, on day one.
Most methods sections are too thin to reproduce — the processing parameters, the software version, the exact phasing are gone. When a result can't be re-derived, it can't be trusted, and increasingly it can't be published.
A tool that prints an answer teaches students nothing. They need to see how an assignment was reached — shift windows, multiplicity, coupling, integration, cross-modal corroboration — to learn to do it themselves.
A shared NMR/MS core serves dozens of labs with wildly different samples. Consistency of processing and QC — not heroics by one expert operator — is what keeps the data comparable across the building.
Academic groups can't absorb per-seat enterprise pricing or a six-month integration project. The tooling has to earn its place by saving time on characterization and SI assembly from day one.
Across the academic workflow
The same evidence object serves every context a department lives in — so what gets taught, characterized, and measured ends up legible in the paper and the deposited dataset too.
The workflow, at a glance
Teaching ──► Bench ──► Core facility ──► Publication ──► Open science
│ │ │ │ │
▼ ▼ ▼ ▼ ▼
see the confirm / consistent QC SI + methods open formats
reasoning elucidate every group auto-composed + provenance
(no black (NMR + MS) (per-peak fit) (each cited) (deposit-ready)
box)TEACH
BENCH
CORE
PUB
OPEN
Workflows we light up
Each is a typed pipeline shipping today. Inputs come from your instruments or your course materials; outputs land in the evidence stack, the SI composer, and the reproducible-methods record.
Confirm a freshly synthesized compound matches its intended structure. NMR + HRMS scored together with DP4 confidence and an evidence trail you can paste into a notebook.
Inputs
Raw FID · mzML · target SMILES
Outputs
Match verdict · DP4 · per-peak assignment
Work a true unknown across 1D/2D NMR and MS/MS. Candidate ranking, fragmentation trees, and cross-modal checks narrow the structure with the reasoning kept visible.
Inputs
1D / 2D NMR · MS/MS · constraints
Outputs
Ranked candidates · connectivity · evidence
Generate a supporting-information package and a methods section from the analysis itself — instrument, solvent, processing recipe, software version, and per-signal assignments, all cited.
Inputs
Reviewed analysis · instrument metadata
Outputs
SI tables · methods draft · recipe hash
Hand a class a real spectrum and let them see the layered reasoning behind each assignment. Transparency is the feature — every category and confidence number traces to its source.
Inputs
Course spectra · candidate structures
Outputs
Layer-by-layer reasoning · answer key
Per-peak fit metrics (χ²ᵣ, RMSE, FWHM, S/N, baseline σ) and identical processing across every user mean the data your facility hands back is comparable group-to-group.
Inputs
Instrument output · processing recipe
Outputs
QC metrics · consistent processed spectra
Methodology groups developing new reactions use Repho to plan screens by Bayesian optimization over yield and selectivity — fewer reactions to find the conditions worth publishing.
Inputs
Reaction recipe · objectives · constraints
Outputs
Next experiment · Pareto front · rationale
Built to be checkable
Each capability below is designed to be re-derived from your own raw data and the shipped release notes. A regression gate runs in CI on every detector change — the kind of rigor you'd expect from a paper, applied to the software.
40
Evidence layers
Typed, additive evidence layers fuse NMR, HRMS, MS/MS, predicted shifts, fragmentation trees, and J-couplings into one transparent confidence score per candidate.
Faster
Dense ¹³C processing
In internal benchmarks (v0.5.0), heavy ¹³C FIDs that previously took several minutes now process in well under a minute — designed to keep a busy shared facility moving.
Auto
Solvent detection
Residual-solvent peaks identified automatically and masked from candidate scoring — designed to reduce manual corrections per spectrum.
Deterministic
Recipe-hash replay
Re-derive a processed spectrum or figure from a prior recipe deterministically — reproducibility a reviewer can verify.
The honest comparison
Most labs run characterization through vendor software, a spreadsheet, and a methods section written from memory. Here's what flips when the evidence — and its provenance — travels with the result.
| Dimension | Today | With MolTrace |
|---|---|---|
| Methods section detail | today Hand-written, often missing processing params and software version | with MolTrace Auto-generated from the analysis with recipe hash, params, and version pinned |
| Reproducing a result later | today A new student re-runs from scratch and hopes it looks the same | with MolTrace Recipe-hash replay re-derives the same output deterministically |
| Teaching how an assignment was reached | today A printout of peaks; the reasoning lives only in the expert's head | with MolTrace Layer-by-layer evidence trail — students see shift, multiplicity, coupling, cross-modal |
| Supporting information assembly | today Manual table-building the week before submission; transcription errors creep in | with MolTrace SI tables composed from the analysis, each value cited to the spectrum it came from |
| Core-facility consistency across users | today Quality depends on which operator processed it that day | with MolTrace Identical processing + per-peak QC for every group — comparable data building-wide |
| Open science & deposition | today Vendor-locked formats; provenance lost when data leaves the instrument PC | with MolTrace Open formats (JCAMP-DX, mzML, CSV) with the immutable raw archive and provenance attached |
One worked example
A single characterization, followed from the instrument to the supporting information — and reproduced by a co-author months later.
Illustrative: a freshly made compound goes in as a raw Bruker FID. SpectraCheck confirms the structure with a DP4 confidence score, and flags a residual CDCl₃ peak at 7.26 ppm so it isn't mistaken for signal.
Every assignment shows its reasoning — shift window, multiplicity, J-coupling, integration, and the HRMS exact mass that corroborates the formula. Nothing is a black-box verdict.
A supporting-information table and a methods paragraph are generated from the analysis — instrument, solvent, processing recipe hash, software version, and per-signal assignments, all cited.
Months later a co-author replays the recipe hash and gets the same output deterministically. The reviewer's reproducibility question answers itself.
Built for trust
The same provenance machinery designed to support pharmaceutical-grade inspection is exactly what an academic group needs for a defensible paper and an honest classroom — open, explainable, and reproducible by anyone who has the data.
Immutable raw archive
Every FID is SHA-256 hashed and never overwritten, so the original evidence behind a published figure survives every reprocess.
Recipe-hash reproducibility
Every processing run links a recipe hash to the unchanged raw archive. Replay a prior recipe and get the same output deterministically.
Transparent, explainable evidence
No black box. Every category, confidence number, and assignment traces to the layer and source that produced it — ideal for teaching and review.
Open formats, no lock-in
JCAMP-DX, mzML, and CSV throughout. Your data stays portable for deposition, collaboration, and the next tool you use.
Contradiction warnings
HRMS exact mass disagreeing with the NMR-implied formula raises a first-class warning before you publish — not after a reviewer finds it.
Your data stays yours
Per-tenant isolation, residency controls designed to support GDPR, and a role-scoped event ledger keep each group's unpublished work private by default.
An unassigned unknown, a teaching example, or a backlog at the core facility — we'll show how MolTrace carries the evidence from the FID to a reproducible figure.