Analytical Method Validation

What it does

Covers the full lifecycle of proving an analytical procedure is fit for its intended purpose across HPLC, LC-MS/MS, GC, CE, ICP-MS, dissolution, qNMR, qPCR, NIR, ligand binding and cell-based assays.

• Framework first: forces a decision between ICH Q2(R2)/Q14, ICH M10, USP <1220>/<1225>/<1226>, CLSI EP and ISO/IEC 17025 so you don't produce a hybrid protocol that satisfies none of them.
• Protocol generation: derives required characteristics and study layout from the measured attribute (assay, quantitative impurity, limit test, identity) and leaves acceptance criteria deliberately blank so they are stated before data collection.
• Correct statistics: lack-of-fit F test and residual runs test instead of bare r²; one-way random-effects model separating repeatability from intermediate precision; DL/QL by every allowed σ approach with confirmation; ICH M10 run acceptance including the per-level 50% rule; TOST equivalence plus Deming and Passing-Bablok regression for method comparison.
• Documentation: protocol and report templates plus reference files on framework selection, Q2(R2) structure, M10 criteria tables, statistics rationale and a source ledger.

Scripts are pure Python 3.11+ standard library, no network, with exit codes 0/1/2 so they can gate a workflow. Importantly, they report but never conclude — declaring a method validated, releasing a batch or closing an OOS stays with the analyst and quality unit.

Who it's for

• QC and analytical development scientists writing validation protocols and reports
• Bioanalysts running PK/TK/BE studies under ICH M10
• Teams transferring a method between labs, sites or instruments who must defend equivalence
• ISO/IEC 17025 accredited labs and regulatory submission authors

Examples

1. Draft an assay protocol: "Build an ICH Q2(R2) validation protocol for an HPLC assay" → returns required characteristics (specificity, response, accuracy, repeatability, intermediate precision), an 80–120% reportable range rationale, and bracketed criteria fields you must fill in first.
2. Catch a bad calibration curve: an r² of 0.983 that would normally pass is rejected once lack-of-fit p = 1.5e-06 and a −9.0% back-calculated error at the low end surface, with a --weight 1/x2 recommendation for heteroscedastic data.
3. Method transfer equivalence: "Are these paired results equivalent within ±2%?" → TOST against a pre-stated margin instead of a non-significant t test, with Deming and Passing-Bablok slopes alongside OLS and a proportional-bias flag.

Install the skill I found on Claude Skill Mart.
Copy the skills/analytical-method-validation folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/analytical-method-validation/.
When it's done, tell me in one line what this skill can do.