interpret-mass-spectrum
정보
이 스킬은 질량 분석 데이터를 분석하여 분자식을 결정하고, 단편화 경로를 식별하며, 구조적 특징을 제안합니다. 개발자들은 이를 사용해 합성 생성물을 확인하거나, 불순물을 식별하거나, 할로겐과 같은 원소의 동위원소 패턴을 해석할 수 있습니다. 이 스킬은 제공된 MS 데이터로부터 분자 이온, 일반적인 단편 손실, 그리고 순도를 체계적으로 평가합니다.
빠른 설치
Claude Code
추천npx 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-mass-spectrumClaude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요
문서
Interpret Mass Spectrum
Analyze MS → mol ion, formula, fragmentation pathways, structural features.
Use When
- MW + formula of unknown
- Confirm synthetic product (mol ion + fragmentation)
- ID impurities / degradation products
- Propose structural features from characteristic frag losses
- Isotope patterns → halogens, S, metals
In
- Req: MS data (m/z + rel int, min full scan)
- Req: Ionization method (EI, ESI, MALDI, CI, APCI, APPI)
- Opt: HRMS exact mass (measured vs calc)
- Opt: Mol formula from other (EA, NMR)
- Opt: MS/MS data (frag of selected precursor)
- Opt: Chrom ctx (LC-MS / GC-MS tR, purity)
Do
Step 1: Ionization Method + Expected Ion Types
Determine what species present before peak assignment:
- Classify ionization:
| Method | Energy | Primary Ion | Fragmentation | Typical Use |
|---|---|---|---|---|
| EI (70 eV) | Hard | M+. (radical cation) | Extensive | Small volatile molecules, GC-MS |
| CI | Soft | [M+H]+, [M+NH4]+ | Minimal | Molecular weight confirmation |
| ESI | Soft | [M+H]+, [M+Na]+, [M-H]- | Minimal | Polar, biomolecules, LC-MS |
| MALDI | Soft | [M+H]+, [M+Na]+, [M+K]+ | Minimal | Large molecules, polymers, proteins |
| APCI | Soft | [M+H]+, [M-H]- | Some | Medium polarity, LC-MS |
- Polarity mode: +ve → cations; -ve → anions. ESI uses both commonly.
- Adducts + clusters: Soft ionization → [M+Na]+ (M+23), [M+K]+ (M+39), [2M+H]+, [2M+Na]+ besides [M+H]+. ID these before mol ion.
- Multiply charged: ESI → m/z = (M + nH) / n. Look for fractional m/z spacing (0.5 Da = z=2).
→ Method documented, expected ion types listed, adducts/clusters ID'd → true mol ion determinable.
If err: Method unknown → examine spectrum for clues: extensive frag → EI; adduct patterns → ESI; matrix peaks → MALDI. Check instrument log.
Step 2: Mol Ion + Mol Formula
ID mol ion peak + derive formula:
- Locate mol ion (M): EI → M+. highest m/z w/ reasonable isotope pattern (may be weak/absent for labile compounds). Soft → ID [M+H]+ / [M+Na]+ + subtract adduct → M.
- N rule: Odd MW → odd # N. Even MW → 0 or even # N.
- DBE: DBE = (2C + 2 + N - H - X) / 2, X = halogens. Ring / π bond = 1 DBE. Benzene = 4, carbonyl = 1.
- HRMS: Exact mass avail → calc formula using mass defect. Compare measured vs candidate formulas in accuracy window (typ < 5 ppm modern instruments).
- Cross-check isotope pattern: Observed must match proposed formula (Step 3).
→ Mol ion ID'd, MW determined, N rule applied, formula proposed (confirmed by HRMS if avail).
If err: No mol ion in EI (common thermally labile / highly branched) → try softer ionization. Ambiguous mol ion → check loss of common small frags from highest m/z (M-1, M-15, M-18 → help ID M).
Step 3: Isotope Patterns
Use isotopic signatures → detect elements:
- Monoisotopic elements: H, C, N, O, F, P, I have characteristic abundances. CHNO only → M+1 ≈ 1.1% per C.
- Halogen patterns:
| Element | Isotopes | M : M+2 Ratio | Visual Pattern |
|---|---|---|---|
| 35Cl / 37Cl | 35, 37 | 3 : 1 | Doublet, 2 Da apart |
| 79Br / 81Br | 79, 81 | 1 : 1 | Equal doublet, 2 Da apart |
| 2 Cl | -- | 9 : 6 : 1 | Triplet |
| 2 Br | -- | 1 : 2 : 1 | Triplet |
| 1 Cl + 1 Br | -- | 3 : 4 : 1 | Characteristic quartet-like |
- Sulfur: 34S → 4.4% at M+2. M+2 ≈ 4-5% rel M (after 13C2 correction) → ≈ 1 S.
- Silicon: 29Si (5.1%) + 30Si (3.4%) → distinctive M+1 + M+2 contributions.
- Compare calc vs observed: Use proposed formula → theoretical pattern → overlay observed → confirm/refute.
→ Pattern analyzed, Cl/Br/S/Si presence determined, consistent w/ proposed formula.
If err: Isotope res insufficient (low-res instrument) → M+2 unresolvable. Note limitation, rely on exact mass + other spectra for elemental comp.
Step 4: Fragmentation Losses + Key Frag Ions
Map pathways → structural info:
- Catalog major frags: All peaks > 5-10% rel int w/ m/z.
- Neutral losses from mol ion:
| Loss (Da) | Neutral Lost | Structural Implication |
|---|---|---|
| 1 | H. | Labile hydrogen |
| 15 | CH3. | Methyl group |
| 17 | OH. | Hydroxyl |
| 18 | H2O | Alcohol, carboxylic acid |
| 27 | HCN | Nitrogen heterocycle, amine |
| 28 | CO or C2H4 | Carbonyl or ethyl |
| 29 | CHO. or C2H5. | Aldehyde or ethyl |
| 31 | OCH3. or CH2OH. | Methoxy or hydroxymethyl |
| 32 | CH3OH | Methyl ester |
| 35/36 | Cl./HCl | Chlorinated compound |
| 44 | CO2 | Carboxylic acid, ester |
| 45 | OC2H5. | Ethoxy |
| 46 | NO2. | Nitro compound |
- Characteristic frag ions:
| m/z | Ion | Origin |
|---|---|---|
| 77 | C6H5+ | Phenyl cation |
| 91 | C7H7+ | Tropylium (benzyl rearrangement) |
| 105 | C6H5CO+ | Benzoyl cation |
| 43 | CH3CO+ or C3H7+ | Acetyl or propyl |
| 57 | C4H9+ or C3H5O+ | tert-Butyl or acrolein |
| 149 | Phthalate fragment | Plasticizer contaminant |
- Map frag pathways: Connect frag ions by successive losses → frag tree from M down to low mass.
- Rearrangement ions: McLafferty (γ-H transfer + β-cleavage) → even-electron ions from carbonyl compounds. Retro-Diels-Alder → characteristic cyclohexene.
→ All major frag ions assigned, neutral losses calc + correlated w/ structure, frag tree built.
If err: Frags don't correspond to simple losses → consider rearrangement. Unassigned frags → unexpected groups, impurities, matrix/BG peaks.
Step 5: Purity + Structure
Evaluate spectrum for purity + assemble proposal:
- Purity check: GC-MS / LC-MS → examine chrom for add'l peaks. Direct-infusion → look for unexpected ions not frags/adducts of analyte.
- BG + contaminant peaks: Common: phthalate plasticizers (m/z 149, 167, 279), column bleed (siloxanes 207, 281, 355, 429 in GC-MS), solvent clusters.
- Structure proposal: Combine formula (Step 2) + isotope (Step 3) + frag (Step 4) → structure / candidate set.
- Rank candidates: Frag tree → rank. Best = explains most frag ions w/ fewest ad hoc.
- Cross-validate: Compare vs NMR, IR, UV-Vis. MS alone rarely unambiguous for novel compounds.
→ Purity assessed, contaminants ID'd if present, structural proposal / ranked candidates consistent w/ all MS + cross-validated where poss.
If err: Multiple components w/o chrom sep → flag mixture, recommend LC-MS / GC-MS reanalysis. No satisfactory proposal → ID which add'l data (HRMS, MS/MS, NMR) would resolve.
Check
- Ionization method ID'd + expected ion types documented
- Mol ion located + distinguished from adducts, frags, clusters
- N rule applied + consistent w/ proposed formula
- DBE calc + accounted for in structure
- Isotope pattern matches formula
- Major frag ions assigned w/ neutral losses + structural rationale
- Frag tree built M → low mass
- Contaminant + BG peaks ID'd + excluded
- Proposal cross-validated w/ other spectra
Traps
- Mis-ID mol ion: EI → base peak often frag, not M. M = highest m/z w/ reasonable isotope pattern. ESI adducts ([M+Na]+, [2M+H]+) → mistaken for M.
- Ignore N rule: Odd-mass M → odd # N. Forget → impossible formulas.
- Confuse isobaric losses: Loss 28 = CO or C2H4; loss 29 = CHO or C2H5. HRMS / add'l frag → distinguish.
- Neglect multiply charged: ESI → 2+/3+ at half/third expected m/z. Non-integer spacing between isotope peaks → multi charge diagnostic.
- Over-interpret low-abundance: Peaks < 1-2% rel int → noise, isotope contribs, minor contaminants, not real frags.
- Assume pure: Many real spectra = mixtures. Check chrom purity + look for ions inconsistent w/ proposed structure.
→
interpret-nmr-spectrum— connectivity + H environments → structural confirminterpret-ir-spectrum— func groups explaining observed fraginterpret-uv-vis-spectrum— chromophores in analyteinterpret-raman-spectrum— complementary vibrationalplan-spectroscopic-analysis— select + sequence techniques pre-acquisitioninterpret-chromatogram— GC/LC chrom data coupled w/ MS
GitHub 저장소
연관 스킬
llamaguard
기타LlamaGuard는 폭력 및 혐오 발언 등 6가지 안전 범주에서 LLM 입력과 출력을 조정하기 위한 Meta의 70-80억 파라미터 모델입니다. 94-95% 정확도를 제공하며 vLLM, Hugging Face 또는 Amazon SageMaker를 사용해 배포할 수 있습니다. 이 기술을 사용하여 AI 애플리케이션에 콘텐츠 필터링 및 안전 가드레일을 손쉽게 통합하세요.
cost-optimization
기타이 Claude Skill은 리소스 적정화, 태깅 전략, 지출 분석을 통해 개발자들이 클라우드 비용을 최적화할 수 있도록 지원합니다. AWS, Azure, GCP에서 클라우드 비용을 절감하고 비용 거버넌스를 구현하기 위한 프레임워크를 제공합니다. 인프라 비용을 분석하거나, 리소스를 적정화하거나, 예산 제약을 충족해야 할 때 사용하세요.
quantizing-models-bitsandbytes
기타이 스킬은 bitsandbytes를 사용하여 LLM을 8비트 또는 4비트 정밀도로 양자화하며, 최소한의 정확도 손실로 50-75%의 메모리 감소를 달성합니다. 제한된 GPU 메모리에서 더 큰 모델을 실행하거나 추론을 가속화하는 데 이상적이며, INT8, NF4, FP4와 같은 형식을 지원합니다. 이 스킬은 HuggingFace Transformers와 통합되어 QLoRA 학습 및 8비트 옵티마이저를 가능하게 합니다.
dispatching-parallel-agents
기타이 Claude Skill은 3개 이상의 독립적인 문제를 동시에 조사하고 해결하기 위해 다중 에이전트를 배치합니다. 공유 상태나 의존성 없이 해결 가능한 무관련 장애 시나리오에 맞게 설계되었습니다. 핵심 기능은 병렬 문제 해결로, 각 독립 문제 영역마다 하나의 에이전트를 할당하여 효율성을 극대화합니다.
