mfsig.com
molforge.ai · signed σ-profile · audit-grade · 84/84 PASS

σ-profiles, the way they should have always been.

Drop in a SMILES, get back a tamper-evident, audit-grade σ-profile. Nice UI. Slick API. Built for FDA submissions, IP filings, and the molecules that have to be right.

try:demo · serves catalogue samples · no compute charged
Already have legacy files?

Drop a vendor σ-profile file — get a signed .mfsig back

Turbomole, COSMOtherm, ORCA, NwChem, Gaussian, xtb, MolForge CSV, and structure formats (XYZ/SDF/MOL). Auto-detect, browser-only, no upload. The full open CLI handles batch + audit hashes.

Drop legacy files here
or click to browse · multiple files OK · stays in your browser, nothing is uploaded
supported:TurbomoleCOSMOthermGaussianORCANwChemxtbMolForge CSV.xyz.sdf.mol
Full converter page →See every .mfsig fieldConversion is free · enrichment is per molecule
What's measured · v0.92.0 production

The numbers that define a signed σ-profile

Every figure below is recomputed from the signed .mfsig files in the production cohort — nothing asserted, everything reproducible. v0.92.0 (wB97M-V / def2-TZVPP) is the production recipe; v0.91.1 (SVP) is the previous validated baseline — see /factory.

+0.9998
σ-profile fidelity (Pearson r) vs CPU reference ground truth · 100-drug cohort
≈ 0
∫σ·dA charge conservation · machine ε on every molecule
84/84
quality gates PASS on the published reference set
SHA-256
tamper-evident per-molecule audit · 21 CFR Part 11 ready
1.1×10⁻⁹
energy fidelity (Hartree) vs reference SCF
< 40 min
100-drug Platinum cohort · same audit trail per molecule
v0.92.0
production recipe (wB97M-V / def2-TZVPP)
~6800
signed .mfsigs in the production atlas

Every .mfsig is generated from quantum mechanics, signed, and reproducible end-to-end. Anatomy of the file: every field, typed and graded. Downstream property predictors built on these photos live on the separate kernels site.

See it live

Open a .mfsig in 3D, right now

Every Pro / Platinum / Reference delivery is a signed .mfsig.json file. Drop one in our 3D viewer and rotate the molecule, inspect the σ-shell, verify the SHA-256 — all in your browser.

v0.22 → v0.28.3

Six things every σ-profile should ship with

Each one shipped, each one measurable, each one in the contract.

v0.22

Five-nines fidelity

σ-profile correlation against the reference CPU ground truth: Pearson r = +0.9998 across the 100-drug cohort. Energy fidelity 1.1 × 10⁻⁹ H.

SHIPPED
v0.23

Conformer-aware

Smart ensemble compression for flexible scaffolds. Boltzmann-weighted σ-profile from the conformers that matter — no manual conformer search.

SHIPPED
v0.24

Universal converter — free forever

Read Turbomole, COSMOtherm, Gaussian, NwChem, ORCA, xtb, MolForge CSV. Convert your legacy archive into the open .mfsig format. No license, no API key.

SHIPPED
v0.25

.mfsig — the open format

SHA-256 audit trail. 21 CFR Part 11 ALCOA-compatible by construction. Reproducibility metadata + lossless source-data vault.

SHIPPED
v0.26

Industrial-scale throughput

100-drug Platinum cohort delivered in under 40 minutes. 389-drug cohorts in 2.5 hours. Same audit trail per drug, every drug.

SHIPPED
v0.27

Analytical cavity

Full Connolly Solvent-Excluded Surface — contact, saddle, and reentrant patches all computed analytically. Five of five unit tests at machine ε.

SHIPPED
v0.27

Platinum geometry

Reference-grade geometry optimisation. Every cohort entry converged to RMS-grad ≤ 10⁻³ H/Bohr. No RDKit-only shortcuts.

SHIPPED
v0.28

Nice UI, slick API

Drop in a SMILES, get back a signed .mfsig.json in seconds. Web UI, REST API, webhook delivery for cohorts — pay only for the molecules you process.

SHIPPED
v0.28.3

Audit-grade σ-profile

Charge conservation at machine ε on every molecule. Hydrogen-bond signal restored across the cohort. 84/84 quality gates PASS on the published 84-molecule reference set.

SHIPPED
v0.29

Three audit tiers

Standard production · Compatibility (drop-in for legacy pipelines) · Premium pure-physics (R&D partner tier with no calibration on the benchmark). Pick the audit context that matches your downstream.

SHIPPED
v2.4

Git-for-σ-profile lineage

Every .mfsig carries a SHA-chained derivation history. Upgrade Pro → Platinum, enrich with extra recipes — all in-place, all signed, all auditable. No re-compute, no double pay.

SHIPPED
Three calibration philosophies

Pick the recipe that matches your audit context

Same audit-grade .mfsig output, three different calibration philosophies — drop-in legacy replacement, COSMO-RS feed, or zero-empirical-training premium R&D.

Every recipe ships the same signed .mfsig schema — switch downstream predictors without touching the σ-profile. Full comparison matrix →

Why audit-grade matters

Four trust signals on every file we ship

Every signed .mfsig you receive carries four independent guarantees — cryptographic, legal, regulatory, and validation. The same file opens in your QA pipeline, your patent attorney's office, and your competitor's verifier — and gives the same answer.

🔐
Ed25519 signed

Every file carries a cryptographic signature. Verify offline with the open mfsig verify CLI. How →

🧬
Reference-validated

σ-profile fidelity r = +0.9998 vs CPU reference ground truth. 84/84 quality gates PASS on the published cohort. Verify →

📜
21 CFR Part 11

ALCOA-compatible audit trail. Append-only lineage chain. Pass to your QA team as-is. Lineage →

🔬
Reproducibility

Every kernel + atlas regeneratable from raw experimental data in <2 days, <$50. Recipe SHA-pinned, vendor SHA-validated on every compute.

What you do with .mfsig

The σ-profile is the foundation — predictions power off it

Once you have an audit-grade σ-profile, any downstream COSMO-RS-family predictor can consume it: solubility, formulation χ, partitioning, phase diagrams, binding. Those predictive kernels live on the separate kernels site.

One signed .mfsig feeds every downstream model — switch the predictor, keep the foundation. The prediction products (solubility, formulation χ, partitioning, phase diagrams, binding) are a separate site, built on exactly these photos.

how it works

Three ways to use it

Web UI for one-offs, REST API for production, webhooks for batch cohorts. Pick whichever fits your workflow.

Web UI — drop in a SMILES

Paste a SMILES, pick a tier, get back a signed .mfsig.json in seconds. No setup.

1.  Paste:  CC(=O)Oc1ccccc1C(=O)O
2.  Pick :  ◉ Pro   ○ Platinum   ○ Reference
3.  Run  :  [ Generate σ-profile ]

Result   :  aspirin.mfsig.json
            SHA-256: e3b0c44298fc1c14...
            σ-profile peak: 25.05
            cavity area: 246.7 Ų
            Pearson r vs ground truth: +0.9998
            [ ⤓ Download ]   [ Plot ]   [ API call ]

REST API — production integration

One POST, get a job ID. Poll or webhook on completion. Audit trail per request.

curl -X POST https://api.mfsig.com/v1/sigma_profile \
  -H "Authorization: Bearer $MFSIG_API_KEY" \
  -d '{
    "smiles": "CC(=O)Oc1ccccc1C(=O)O",
    "tier":   "pro",
    "solvent":"water",
    "webhook_url": "https://your.app/cb"
  }'

# → { "job_id": "job_8K2…", "status": "queued" }

Cohort delivery — batch + webhook

Submit a SMILES list, receive .mfsig.tar.gz when done. Same audit per drug.

curl -X POST https://api.mfsig.com/v1/cohort \
  -H "Authorization: Bearer $MFSIG_API_KEY" \
  -F smiles=@my_drugs.txt \
  -F tier=platinum \
  -F webhook_url=https://your.app/cb

# → returns { "cohort_id": "co_3Y…",
#             "n_molecules": 389,
#             "eta_min": 158 }
# → delivers a signed cohort.tar.gz on completion

Free converter — legacy migration

Convert your existing Turbomole / COSMOtherm / Gaussian archive into the open .mfsig format.

# Drop your legacy file at https://mfsig.com/converter

# Or use the open CLI — free, no key
mfsig convert legacy_run.cosmo  -o out.mfsig.json
mfsig convert /data/legacy/     -o /data/mfsig/   --batch

# All seven vendor formats auto-detected:
# Turbomole · COSMOtherm · Gaussian · NwChem
# ORCA · xtb · MolForge CSV
Three steps, sixty seconds

Try it now

Drop a SMILES in the web demo. Get back a signed .mfsig.json in seconds. No setup, no install.

# 1.  Sign in at https://mfsig.com

# 2.  Drop a SMILES in the web UI, pick a tier, run.
#     Or, the same flow from your terminal:

$ curl -X POST https://api.mfsig.com/v1/sigma_profile \
       -H "Authorization: Bearer $MFSIG_API_KEY" \
       -d '{"smiles":"CC(=O)Oc1ccccc1C(=O)O","tier":"pro"}'

# 3.  We deliver a signed .mfsig.json — yours to keep,
#     verifiable for life by anyone who has the file.