nmr-structure-elucidation
/nmr-structure-elucidation
Rozwiązuje struktury małych cząsteczek na podstawie widm NMR, MS i IR oraz sprawdza przypisania sygnałów i konfigurację względną.
- Autor
- Soft1919 ↗
- Repozytorium
- NMR-structure-elucidation ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 28.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 5
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 28.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Rozwiązuje struktury małych cząsteczek na podstawie widm NMR, MS i IR oraz sprawdza przypisania sygnałów i konfigurację względną.
Oryginalny opis (EN)
Solve small-molecule structure elucidation problems from 1D/2D NMR (1H, 13C, DEPT, HSQC, COSY, HMBC, NOESY) plus MS/IR, including spectra embedded as vector images in Word/PowerPoint files. Use when the user gives spectra and a formula and asks for the structure, the relative configuration, a signal assignment, or a check of a proposed/published structure.
Jak działa
- 0. Get exact numbers
- 1. Remove what is not the compound
- 2. Inventory
- 3. Constitution
- 4. Relative configuration
- 5. Report
- Pitfalls
Podgląd SKILL.md
--- name: nmr-structure-elucidation description: Solve small-molecule structure elucidation problems from 1D/2D NMR (1H, 13C, DEPT, HSQC, COSY, HMBC, NOESY) plus MS/IR, including spectra embedded as vector images in Word/PowerPoint files. Use when the user gives spectra and a formula and asks for the structure, the relative configuration, a signal assignment, or a check of a proposed/published structure. --- # NMR structure elucidation Work from the spectra. Do not look up the answer or try to identify the compound in the literature unless the user asks; a published structure can be wrong, and the deliverable is a structure that the data support. State clearly what is proven, what is inferred, and what is open. ## 0. Get exact numbers - Spectra inside .docx/.pptx are often EMF vector graphics. Extract them instead of reading pictures: `python scripts/emf.py INPUT.docx --out work/emf`, then `python scripts/spectra.py peaks1d|mult|integ|peaks2d|proj|zoom ...` (see `--help`). - Check the printed calibration residuals. Note the digital resolution of each trace (decimated traces merge close peaks and under-integrate sharp singlets). - 2D plots usually carry their own 1D projections (`proj`): a second, independent peak list and a check that 1D and 2D data come from the same sample. - Raster-only spectra: read them, but quote the reading uncertainty. ## 1. Remove what is not the compound Before interpreting, list and exclude: solvent and labelled impurities; residual CH2Cl2 / grease / water; HMBC one-bond artefacts (pairs symmetric about a 1H shift, split by 1J(CH) ~125-180 Hz, sitting on that proton's own HSQC carbon); DEPT peaks with no 13C counterpart; NOESY cross-peaks with the same sign as the diagonal (exchange/COSY artefacts in small molecules). ## 2. Inventory 1. Formula -> degrees of unsaturation; nitrogen rule; IR functional groups. 2. Every carbon: 13C + DEPT-135 sign + HSQC protons -> C / CH / CH2 / CH3. Overlapped carbons show up as mixed DEPT signs or extra HSQC protons; the carbon count must equal the formula. 3. Count C-bound H; the difference to the formula = OH/NH. Distribute heteroatoms (shift ranges) and the remaining unsaturation (C=X vs rings). ## 3. Constitution 1. COSY + HSQC -> spin fragments; note where each fragment ends (quaternary C, heteroatom). 2. HMBC joins fragments. Use only 2J/3J; methyl singlets are the most reliable anchors. 3. Use valence bookkeeping: when only two atoms have open valences, they must be bonded. 4. Write down every competing constitution that fits the formula and class of compound. Test all of them with `scripts/topology.py` (exhaustive assignment search; reports the minimum number of violated HMBC/COSY correlations). Zero violations is required for the proposed structure; every violation of a rival must be a real, checked correlation. 5. Look for correlations that must exist for a candidate and are absent (e.g. an expected large vicinal coupling or COSY cross-peak). Absences are often decisive. ## 4. Relative configuration 1. Qualitative first: large 3J (>8 Hz) = near-antiperiplanar; NOE contacts within ~3 A. Rigid polycycles: decide which face each substituent occupies. 2. Then exhaustive: `scripts/stereo.py CONFIG.json` enumerates stereoisomers, discards non-embeddable ones, Boltzmann-averages MMFF conformers and scores HLA-Karplus J values and r^-6 distances against the data. Check robustness: re-weight NOE vs J, drop one observable at a time; report which observable decides each centre. 3. NMR gives relative configuration only. Never state an absolute configuration from NMR. ## 5. Report - Final structure (SMILES with stereo + a drawing), full assignment table (dC, type, dH (mult, J), key HMBC, key NOE), and the argument chain step by step. - A short "not proven" list: interchangeable assignments, overlapped signals, configurations that rest on a single observable, absolute configuration. - If a reference/published structure disagrees, list the specific correlations it violates and the experiment that would settle it. Do not adopt it silently. ## Pitfalls - Drawing: build 2D coordinates yourself and re-read the molblock to confirm stereo; removing H atoms one by one can corrupt parity (use RemoveHs). Label substituent carbons too. - Karplus values carry ~1-2 Hz model error; compare isomers with one consistent method. - Do not trust a single weak cross-peak, and do not ignore a strong one. - Never redistribute the source documents or embedded OLE/MestReNova objects: they can contain user names, file paths and sample IDs. Strip them before sharing anything. See `references/checklist.md` for a printable checklist and the config formats.
Pliki w paczce (5)
- .gitignore 1 KB
- LICENSE 1 KB
- README.md 2 KB
- SKILL.md 5 KB
- requirements.txt 1 KB
Wersje
W bazie nie ma innych skilli o nazwie nmr-structure-elucidation.
Migawka bazy z 13.11.2026 · wersja paczki c7cf1e23923d (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 90si-nmr-review-guide
Ocenia widma 1H NMR w materiałach uzupełniających publikacji chemicznej według wytycznych grupy, nadając ocenę E, V, S lub N z uzasadnieniem. - 100si-nmr-review
Wstępnie ocenia widma 1H NMR z pliku SI publikacji chemicznej według wytycznych grupy, z propozycją oceny dla recenzenta. - 100alterlab-hmdb
Przeszukuje bazę ludzkiego metabolomu HMDB (ponad 220 tys. metabolitów) po nazwie, identyfikatorze lub strukturze, pobierając właściwości chemiczne, biomarkery, widma NMR i MS oraz szlaki. - 100nmrglue
Przetwarza surowe zaniki swobodnej indukcji NMR na widma: fazowanie, wykrywanie pików, całkowanie i korekcję linii bazowej. - 100ml-nmr-methodology
Dostarcza metodykę ML-NMR: integrację danych indywidualnych (IPD) i zagregowanych (AgD), korektę populacji, całkowanie numeryczne i predykcję dla populacji docelowych, przy prowadzeniu i recenzji analiz. - 100tooluniverse-organic-chemistry
Pomaga rozwiązywać zadania z chemii organicznej: przewiduje produkty reakcji, analizuje mechanizmy i interpretuje widma NMR, IR oraz MS, rozumując od zasad pierwszych.