validate-analytical-method
О программе
Этот навык Claude проверяет хроматографические аналитические методы в соответствии с регуляторными руководствами ICH Q2(R2). Он проводит разработчиков через определение области валидации, установление специфичности с помощью вынужденной деградации и определение линейности, правильности, прецизионности и пределов обнаружения. Используйте этот навык для структурирования и документирования полной валидации метода для подачи в регуляторные органы.
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/validate-analytical-methodСкопируйте и вставьте эту команду в Claude Code для установки этого навыка
Документация
Validate Analytical Method
Formal valid of chromatographic method per ICH Q2(R2) → scope by category → specificity via forced degradation → linearity + range → accuracy + precision → LOD/LOQ + robustness → regulatory compliance.
Use When
- New chromatographic method developed → must validate before routine
- Compendial method verified for suitability in specific lab
- Validated method significant changes → partial|full re-valid
- Prep validation pkg for regulatory submission (NDA, ANDA, MAA, IND)
- Transferring method to new lab|instrument platform
In
Required
- Developed method: Fully optimized + documented (col, mobile phase, gradient, detector, etc.)
- Method category: Assay active, quant impurity, limit impurity, ID test
- Analyte ref stds: Primary refs w/ COA + assigned purity
- Sample matrix: Representative samples + placebo|blank for specificity
Optional
- Regulatory guidance: Beyond ICH Q2 (USP <1225>, FDA, EMA)
- Forced degradation samples: Pre-stressed (acid, base, ox, heat, light) if not prepared
- Validation protocol: Pre-approved w/ acceptance criteria (req in GMP)
- Transfer pkg: If method transfer → originating lab's report
Do
Step 1: Define Scope per ICH Q2(R2)
ID method category + required validation params.
| Parameter | Cat I: Assay | Cat II: Impurity (Quant) | Cat III: Impurity (Limit) | Cat IV: Identification |
|---|---|---|---|---|
| Specificity | Yes | Yes | Yes | Yes |
| Linearity | Yes | Yes | No | No |
| Range | Yes | Yes | No | No |
| Accuracy | Yes | Yes | No | No |
| Precision (repeatability) | Yes | Yes | No | No |
| Precision (intermediate) | Yes | Yes | No | No |
| LOD | No | May be needed | Yes | No |
| LOQ | No | Yes | No | No |
| Robustness | Yes | Yes | Yes | No |
- Classify method → 4 ICH categories per intended purpose
- Per table, ID required params
- Define acceptance criteria per param before exp work. Typical:
- Linearity: R² >= 0.999 (assay), >= 0.99 (impurity)
- Accuracy: recovery 98.0-102.0% (assay), 80-120% at LOQ
- Repeatability: RSD <= 2.0% (assay), <= 10% at LOQ
- Intermediate: RSD <= 3.0% (assay)
- Draft protocol w/ all params, designs, criteria
- GMP → protocol approval before exp work
Got: Approved protocol w/ method category, required params, designs, pre-defined criteria.
If err: Method category ambiguous (combined assay + impurity) → validate for most stringent. Consult regulatory guidance per submission type.
Step 2: Specificity + Selectivity
- Prep solutions:
- Blank (solvent|diluent)
- Placebo (matrix w/o analyte, e.g., excipients)
- Ref std at working conc
- Spiked placebo (matrix + ref)
- Forced degradation samples (if not avail)
- Forced degradation → potential degradation products:
| Stress Condition | Typical Treatment | Target Degradation |
|---|---|---|
| Acid hydrolysis | 0.1-1 N HCl, 60-80 C, 1-24 h | 5-20% |
| Base hydrolysis | 0.1-1 N NaOH, 60-80 C, 1-24 h | 5-20% |
| Oxidation | 0.3-3% H2O2, RT-60 C, 1-24 h | 5-20% |
| Thermal | 60-80 C, solid state, 1-7 days | 5-20% |
| Photolytic | ICH Q1B (1.2M lux-hours, 200 Wh/m^2 UV) | 5-20% |
- Inject all + eval:
- No interfering peaks from blank|placebo at analyte RT
- Degradation products resolved from main analyte (Rs >= 1.5)
- Peak purity → DAD spectral purity index|MS
- Mass balance: assay + impurities + degradation = 95-105% initial
- Doc specificity w/ chromatograms from all conditions
Got: Method specific: no interferences, degradation resolved, peak purity confirmed, mass balance 95-105%.
If err: Degradation co-elutes w/ analyte → method not stability-indicating. Return to method dev → improve selectivity (pH, gradient, col chem) before validation.
Step 3: Linearity + Range
- 5+ conc levels spanning intended range:
- Assay: 80-120% of target
- Impurity: LOQ → 120-200% of spec limit
- Dissolution: 10-120% of label claim
- Each conc independent (not serial dilution) → best practice
- Inject each in triplicate (min duplicate)
- Linear regression of response (area|height) vs conc:
- Report slope, intercept, R²
- R² >= 0.999 assay; >= 0.99 impurity
- Eval residual plots:
- Random distribution around zero, no systematic pattern
- Curved → non-linearity → quadratic fit or narrower range
- Calc y-intercept as % of response at 100%:
- <= 2% of 100% response for assay
- Validated range = interval between low + high concs w/ linearity, accuracy, precision
Got: Linear regression R² >= 0.999 (assay)|>= 0.99 (impurity), random residuals, intercept <= 2% target, range defined.
If err: R² below criterion → check prep errors, detector non-linearity (too high conc), analyte instability. Repeat w/ fresh prep. Inherent non-linearity → polynomial calibration|narrow range.
Step 4: Accuracy
- Prep accuracy samples at 3 conc (80%, 100%, 120% target for assay; LOQ, mid, high for impurity)
- Per level → 3 independent replicates (min 9 determinations total)
- Drug substance: compare found vs known (gravimetric)
- Drug product: spiked placebo → add known to placebo + measure recovery
- % recovery per level:
- Recovery (%) = (found / added) x 100
- Acceptance:
| Method Type | Recovery Range | RSD at Each Level |
|---|---|---|
| Assay (drug substance) | 98.0-102.0% | <= 2.0% |
| Assay (drug product) | 98.0-102.0% | <= 2.0% |
| Impurity (quantitation) | 80-120% at LOQ, 90-110% at higher levels | <= 10% at LOQ, <= 5% at higher |
| Cleaning validation | 70-130% (or tighter per company SOP) | <= 15% |
- Report individual recoveries, mean, RSD per level
Got: Mean recovery within criteria all levels, RSD within limits.
If err: Recovery consistently high|low all levels → systematic err in ref std, sample prep, method (matrix effect → ion suppression in LC-MS). Erratic → investigate prep technique + analyte stability.
Step 5: Precision
3 levels:
- Repeatability (intra-day):
- 1 analyst, 1 instrument, 1 day
- 6 determinations at 100% or 3 levels x 3 replicates (same as accuracy)
- RSD <= 2.0% assay, <= 10% at LOQ for impurity
- Intermediate (inter-day|inter-analyst):
- Repeat repeatability w/ different analyst, day, (if avail) instrument
- Overall RSD combining both
- Overall RSD <= 3.0% assay
- Intermediate >> repeatability → investigate source (analyst, instrument cal, env)
- Reproducibility (method transfer|multi-site):
- Receiving lab → same protocol
- Compare across labs
- F-test (variance) + t-test (mean) or equivalence
| Precision Level | Design | Acceptance (Assay) | Acceptance (Impurity Quant) |
|---|---|---|---|
| Repeatability | n >= 6 at 100%, 1 analyst, 1 day | RSD <= 2.0% | RSD <= 10% at LOQ, <= 5% above |
| Intermediate | 2 analysts, 2 days (or 2 instruments) | RSD <= 3.0% | RSD <= 15% at LOQ, <= 10% above |
| Reproducibility | Multi-laboratory | Per protocol / transfer criteria | Per protocol / transfer criteria |
Got: Repeatability + intermediate RSDs within criteria. No statistically significant diff between analysts|days|instruments beyond allowed RSD.
If err: Intermediate >> repeatability → ID variable driving variance (analyst prep, ambient temp, instrument cal drift) → control before repeating.
Step 6: LOD, LOQ, Robustness
LOD + LOQ:
- Calc via S/N or std dev approach:
- LOD = 3.3 x (σ / S) → σ = SD of response at low conc, S = slope of cal
- LOQ = 10 x (σ / S)
- Alt: S/N → LOD = S/N >= 3, LOQ = S/N >= 10
- Confirm experimentally → prep solutions at calc'd LOD + LOQ + inject:
- LOD: detectable, not necessarily quantifiable w/ acceptable precision
- LOQ: 6 replicates → RSD <= 10% + accuracy 80-120%
- Report LOD + LOQ w/ method used
Robustness:
- ID critical params (5-7 factors):
- Mobile phase comp (+/- 2% organic)
- MP pH (+/- 0.2 units)
- Col temp (+/- 5 C)
- Flow rate (+/- 10%)
- Detection wavelength (+/- 2 nm)
- Col lot|batch (if avail)
- Vary each within range while holding others (or fractional factorial design for efficiency)
- Eval impact on sys suitability (RT, resolution, tailing, area)
- Params causing sys suitability fail w/in tested range → tightly control + doc as critical
- Summarize robustness in table: each varied param, range tested, impact on responses
Got: LOD + LOQ experimentally confirmed. Robustness done w/ critical params ID'd + control limits.
If err: LOQ precision > 10% RSD → method sensitivity insufficient. Options: increase inj volume, concentrate sample, improve cleanup, more sensitive detector. Param shows not robust (fails SST w/ small variation) → tighten control + flag during transfer.
Check
- Method category ID'd + all params per ICH Q2(R2)
- Validation protocol w/ pre-defined criteria
- Specificity demonstrated: no interferences, degradation resolved, peak purity
- Mass balance 95-105% for forced degradation
- Linearity R² >= 0.999 (assay)|>= 0.99 (impurity), residuals random
- Accuracy at 3 levels, recovery within criteria
- Repeatability RSD within limits (<= 2.0% assay)
- Intermediate RSD within limits (<= 3.0% assay)
- LOD + LOQ experimentally confirmed (LOQ precision <= 10% RSD)
- Robustness w/ critical params ID'd
- All raw data, calc, chromatograms in validation report
Traps
- Exp before protocol approval: GMP → data before protocol approval may not be acceptable to regulators. Always approve first.
- Serial dilutions for linearity: Propagates pipetting errs. Each conc independent from common stock for accurate linearity.
- Insufficient forced degradation: < 5% misses important products. > 30% → secondary products complicate. Target 5-20% per condition.
- Confuse repeatability + intermediate: Repeatability = same-day, same-analyst, same-instrument. Intermediate must vary one. Both required Cat I + II.
- Neglect LOQ verify step: Calc LOQ from cal curve insufficient. Must experimentally confirm acceptable precision + accuracy.
- Robustness late: Discovering method not robust after accuracy + precision wastes time + materials. Quick screen early → catch fragile params.
- Incomplete reports: Reviewers expect all raw data, chromatograms (not just tabulated nums), stat analysis, explicit pass/fail per param. Missing → deficiency letters.
→
develop-gc-method— GC method dev preceding validationdevelop-hplc-method— HPLC method dev preceding validationinterpret-chromatogram— reading chromatograms during validationtroubleshoot-separation— resolve issues during validationconduct-gxp-audit— auditing completed validation for GxPwrite-standard-operating-procedure— doc validated method as SOP
GitHub репозиторий
Похожие навыки
evaluating-llms-harness
ТестированиеЭтот навык Claude запускает lm-evaluation-harness для тестирования LLM на более чем 60 стандартизированных академических задачах, таких как MMLU и GSM8K. Он предназначен для разработчиков, чтобы сравнивать качество моделей, отслеживать прогресс обучения или сообщать академические результаты. Инструмент поддерживает различные бэкенды, включая модели HuggingFace и vLLM.
cloudflare-cron-triggers
ТестированиеЭтот навык предоставляет обширные знания по реализации Cloudflare Cron Triggers для планирования запуска Workers с помощью cron-выражений. Он охватывает настройку периодических задач, заданий технического обслуживания и автоматизированных рабочих процессов, а также решение распространенных проблем, таких как неверные cron-выражения и ошибки часовых поясов. Разработчики могут использовать его для настройки планировщиков обработчиков, тестирования cron-триггеров и интеграции с Workflows и Green Compute.
webapp-testing
ТестированиеЭтот навык Claude предоставляет инструментарий на базе Playwright для тестирования локальных веб-приложений с помощью Python-скриптов. Он позволяет проводить проверку фронтенда, отладку интерфейса, создание скриншотов и просмотр логов, одновременно управляя жизненным циклом сервера. Используйте его для задач автоматизации браузера, но запускайте скрипты напрямую, вместо чтения их исходного кода, чтобы избежать загрязнения контекста.
finishing-a-development-branch
ТестированиеЭтот навык помогает разработчикам завершать готовую работу, проверяя прохождение тестов и предлагая структурированные варианты интеграции. Он направляет рабочий процесс по слиянию, созданию пул-реквестов или очистке веток после завершения реализации. Используйте его, когда ваш код готов и протестирован, чтобы систематически завершать процесс разработки.
