interpret-nmr-spectrum
О программе
Этот навык анализирует одномерные и двумерные данные ЯМР (такие как ¹H, ¹³C, COSY, HSQC) для присвоения химических сдвигов, интерпретации картины спин-спинового взаимодействия и интеграции многомерных корреляций. Он используется для определения неизвестных молекулярных структур или подтверждения синтезированных продуктов путём предложения непротиворечивых структурных фрагментов. Разработчики могут применять его для систематической интерпретации спектров при работе со сложными, перекрывающимися данными.
Быстрая установка
Claude Code
Рекомендуетсяnpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/interpret-nmr-spectrumСкопируйте и вставьте эту команду в Claude Code для установки этого навыка
Документация
Interpret NMR Spectrum
Analyze 1D + 2D NMR → assign peaks, determine coupling, propose structural fragments consistent w/ all data.
Use When
- Structure of unknown organic compound from NMR
- Confirm identity + purity of synthesized product
- Assign peaks in complex spectra w/ overlap
- Correlate multi-exp (1H, 13C, DEPT, COSY, HSQC, HMBC) → unified picture
- Distinguish regioisomers / stereoisomers / conformational
In
- Req: NMR data (min 1H w/ shifts, multiplicities, integration)
- Req: Mol formula / MW (from MS, EA)
- Opt: 13C + DEPT (shifts + multiplicities)
- Opt: 2D (COSY, HSQC, HMBC, NOESY/ROESY correlation tables)
- Opt: Solvent + field strength
- Opt: Known constraints (rxn starting material, IR confirmed groups)
Do
Step 1: Spectrum Type + Acquisition
Establish what data + quality before interpret:
- ID exp types: Catalog which avail (1H, 13C, DEPT-135, DEPT-90, COSY, HSQC, HMBC, NOESY, ROESY, TOCSY). Note nucleus + dimensionality.
- Acquisition params: Spectrometer freq (400 MHz, 600 MHz), solvent, temp, ref standard.
- Solvent + ref peaks: Locate + exclude.
| Solvent | 1H Residual (ppm) | 13C Signal (ppm) |
|---|---|---|
| CDCl3 | 7.26 | 77.16 |
| DMSO-d6 | 2.50 | 39.52 |
| D2O | 4.79 | -- |
| CD3OD | 3.31 | 49.00 |
| Acetone-d6 | 2.05 | 29.84, 206.26 |
| C6D6 | 7.16 | 128.06 |
- Quality: Baseline flatness, multiplet res, S/N. Flag artifacts (spinning sidebands, 13C satellites, solvent impurity H2O ~1.56 ppm CDCl3).
→ Inventory of exps, solvent/ref peaks excluded, quality assessed.
If err: Poor S/N / severe baseline distortion → note limitation + cautious. Flag peaks indistinguishable from noise.
Step 2: 1H Chemical Shifts
Assign each 1H → environment using shift ranges:
- Tabulate: Per peak → shift (ppm), multiplicity, J (Hz), rel int.
- Classify by shift:
| Range (ppm) | Environment | Examples |
|---|---|---|
| 0.0--0.5 | Shielded (cyclopropane, M-H) | Cyclopropyl H, metal hydrides |
| 0.5--2.0 | Alkyl (CH3, CH2, CH) | Saturated aliphatic chains |
| 2.0--4.5 | Alpha to heteroatom/unsaturation | -OCH3, -NCH2, allylic, benzylic |
| 4.5--6.5 | Vinyl / olefinic | =CH-, =CH2 |
| 6.5--8.5 | Aromatic | ArH |
| 9.0--10.0 | Aldehyde | -CHO |
| 10.0--12.0 | Carboxylic acid | -COOH |
| 0.5--5.0 (broad, exchangeable) | OH, NH | Alcohols, amines, amides |
- Count H: Integration ratios rel to formula → # protons per signal. Normalize simplest whole-# ratio.
- Exchangeable protons: Signals disappear on D2O shake (OH, NH, COOH) = exchangeable. Record presence + shift.
→ Table of 1H signals w/ shift, multiplicity, J, integration (# H), prelim env assignment.
If err: Integration doesn't sum to expected → check overlapping, broad peaks hidden in baseline, wrong formula.
Step 3: Coupling Patterns + J-Values
Extract connectivity from splitting:
- Multiplicities: s, d, t, q, dd, etc. Complex m → estimate # coupling partners.
- Measure J: Extract Hz. Match reciprocal (if H_A ↔ H_B J = 7.2, H_B shows same J to H_A).
- Classify J:
| J Range (Hz) | Coupling Type |
|---|---|
| 0--3 | Geminal (2J) or long-range (4J, 5J) |
| 6--8 | Vicinal aliphatic (3J) |
| 8--10 | Vicinal with restricted rotation |
| 10--17 | Vicinal olefinic cis (6--12) or trans (12--18) |
| 0--3 | Aromatic meta |
| 6--9 | Aromatic ortho |
- Map coupling networks: Group mutually coupled protons → spin systems. Each = connected frag.
- Roof effect: AB-type → inner lines of doublets more intense → chemical shift proximity.
→ All J measured + matched reciprocally, spin systems ID'd, coupling types classified.
If err: Multiplets too complex for first-order → note higher-order pattern. Overlapping / strongly coupled (δν/J < 10) → non-first-order requires simulation.
Step 4: 13C + DEPT
Determine C types + count:
- Count distinct 13C signals: Compare # peaks vs formula. Fewer → symmetry.
- Classify by shift:
| Range (ppm) | Carbon Type | Examples |
|---|---|---|
| 0--50 | sp3 Alkyl | CH3, CH2, CH, quaternary C |
| 50--100 | Alpha to O or N | -OCH3, -OCH2, anomeric C |
| 100--150 | Aromatic / vinyl | =CH-, ArC |
| 150--170 | Heteroaromatic / enol / imine | C=N, C-O aromatic |
| 170--185 | Carboxyl / ester / amide | -COOH, -COOR, -CONR2 |
| 185--220 | Aldehyde / ketone | -CHO, >C=O |
- DEPT editing: DEPT-135 (CH + CH3 up, CH2 down, quaternary absent) + DEPT-90 (CH only) → # attached H per C.
- DBE: DBE = (2C + 2 + N - H - X) / 2. Compare # π bonds + rings implied.
→ Every 13C signal classified by type (CH3, CH2, CH, C) + env, DBE consistent w/ observed groups.
If err: No DEPT → infer H attachment from HSQC (Step 5). C count ≠ formula → coincident signals / quaternary Cs in noise.
Step 5: 2D NMR
Build connectivity using 2D exps:
- COSY (1H-1H): Which H 2-3 bonds apart. Map cross-peaks → confirm+extend spin systems Step 3.
- HSQC (1H-13C 1-bond): Assign each H → directly bonded C. Links 1H + 13C unambiguously.
- HMBC (1H-13C long-range): 2-3 bond H-C. Critical for connecting frags across quaternary C, heteroatoms, carbonyls w/o direct H-C.
- NOESY/ROESY (through-space): H's spatially close (<5 Å) regardless bonding. → Stereochem + conformational.
- Build frag connectivity: HMBC → connect COSY spin systems → larger frags. Each HMBC cross-peak = 2-3 bond H-C path.
→ Connectivity map linking spin systems into coherent framework + stereochem from NOE where avail.
If err: 2D incomplete / ambiguous → note tentative connections. Multiple proposals poss. Prioritize HMBC → bridges gaps COSY can't.
Step 6: Propose + Validate Structure
Assemble frags → complete proposal:
- Assemble: Connect frags Steps 2-5 using HMBC + DBE constraints.
- Check formula: Proposed matches formula exactly (atom count, DBE).
- Back-predict shifts: For proposed → predict 1H + 13C shifts. Compare observed; deviations > 0.3 ppm (1H) / > 5 ppm (13C) → re-examine.
- Verify all correlations: Every COSY, HSQC, HMBC explained. Unexplained → error / impurity.
- Alternatives: Multiple structures fit → list distinguishing exps / correlations.
- Stereochem: NOE + J analysis (Karplus for dihedral) + known conformational prefs → relative + (where poss) absolute.
→ Single best-fit proposal w/ all NMR accounted, or ranked candidates + plan to distinguish.
If err: No single structure → check: mixture (extra peaks non-integer int), dynamic processes (broad peaks from conformational exchange), paramagnetic impurities (anomalous broadening). Re-examine formula if multiple equally viable.
Check
- Solvent + ref peaks ID'd + excluded
- Every 1H signal → shift region, multiplicity, J, integration
- J reciprocal (matched between partners)
- 13C classified by DEPT multiplicity + shift
- DBE calc + consistent w/ proposed
- 2D (COSY, HSQC, HMBC) all explained
- Proposed matches formula exactly
- Back-predicted shifts agree w/ observed within tolerance
- Stereochem via NOE / J where applicable
Traps
- Ignore solvent peaks: Common solvents → signals overlap analyte. Always ID + exclude residuals, water, grease.
- Force 1st-order on 2nd-order: Strongly coupled nuclei (small Δshift rel J) → distorted multiplets, can't interpret w/ simple n+1. Roof effects + non-binomial intensity → indicators.
- Overlook exchangeable: OH + NH may be broad, shift w/ conc/temp, absent in protic solvents. D2O shake → clarifies.
- Assume all 13C visible: Quaternary Cs → long relax times + low int. May be absent short-acquisition. HMBC often only way to detect.
- Misinterpret HMBC artifacts: HMBC → 1-bond artifacts (mis-assigned long-range) + weak 4-bond. Cross-check w/ HSQC → filter 1-bond leakthrough.
- Neglect symmetry: Fewer 13C peaks than formula → symmetry element. Account before proposing.
→
interpret-ir-spectrum— func groups → constrain NMR structureinterpret-mass-spectrum— formula + frag for cross-valinterpret-uv-vis-spectrum— chromophores + conjugation extentinterpret-raman-spectrum— complementary vibrational → symmetric modesplan-spectroscopic-analysis— select + sequence techniques pre-acquisition
GitHub репозиторий
Похожие навыки
llamaguard
ДругоеLlamaGuard — это модель от Meta с 7–8 миллиардами параметров для модерации входных и выходных данных больших языковых моделей по шести категориям безопасности, таким как насилие и разжигание ненависти. Она обеспечивает точность 94–95% и может быть развернута с помощью vLLM, Hugging Face или Amazon SageMaker. Используйте этот навык, чтобы легко интегрировать фильтрацию контента и защитные механизмы в ваши ИИ-приложения.
cost-optimization
ДругоеЭтот навык Claude помогает разработчикам оптимизировать облачные расходы за счет правильного подбора ресурсов, стратегий тегирования и анализа затрат. Он предоставляет framework для сокращения облачных расходов и внедрения управления затратами в AWS, Azure и GCP. Используйте его, когда вам нужно проанализировать расходы на инфраструктуру, оптимизировать ресурсы или уложиться в бюджетные ограничения.
quantizing-models-bitsandbytes
ДругоеЭтот навык выполняет квантизацию LLM до 8-битной или 4-битной точности с использованием библиотеки bitsandbytes, обеспечивая сокращение использования памяти на 50-75% при минимальной потере точности. Он идеально подходит для запуска больших моделей при ограниченной памяти GPU или для ускорения вывода, поддерживая форматы INT8, NF4 и FP4. Навык интегрируется с HuggingFace Transformers и позволяет использовать обучение QLoRA и 8-битные оптимизаторы.
dispatching-parallel-agents
ДругоеЭтот навык Claude распределяет нескольких агентов для исследования и устранения трёх и более независимых проблем параллельно. Он предназначен для сценариев с несвязанными сбоями, которые можно устранить без общего состояния или зависимостей. Ключевая возможность — параллельное решение проблем, где за каждую независимую предметную область назначается отдельный агент для максимальной эффективности.
