# Deep-dive research — Round 2 **Last updated:** 2026-04-23 (corrections applied after round-2 red-team critique — see [`../../review/critique-round2.md`](../../review/critique-round2.md) for the full list; notable fixes: USP <621> LC method-adjustment limits in DD1, tesamorelin MW in DD3, CJC-1295 Leu²⁷ vs Met²⁷, DD1/DD5 temperature standardisation on 30 °C, Artemis Analytical newly surfaced public website, THR test count 16 not 18, ACC catalog W0051-10 vs W055A). **Compiled by an AI agent** using the structured claim format in `/workspace/overview/RESEARCH-BEST-PRACTICES.md`. Numeric claims carry C1–C5 confidence tiers; unflagged prose is inference and should be treated as C3 or worse. Round 1 shipped the breadth corpus (~86,000 words, 60 files — entry point: [`../summary.md`](../summary.md)). Round 2 is **depth of execution**: protocols a reader could actually run on a refurb Agilent 1100 or an ACC Pyrotell kit, profiles of the community labs that do this work today, and the aggregate statistics that describe the ecosystem. Total Round 2 corpus: ~53,000 words across 8 deliverables + 11 lab profiles + this index. --- ## The 8 deliverables ### DD1 — [`sop-estradiol-valerate-hplc.md`](sop-estradiol-valerate-hplc.md) (5,647 words) Replicable HPLC SOP for estradiol valerate in oil on an Agilent 1100. Covers module selection (G1311A vs G1312A, G1315B DAD vs VWD), ChemStation B.04.03 revision + 2026-01-31 EOL, two verified columns (Waters Symmetry C18 WAT045905; Agilent ZORBAX Eclipse Plus C18 959993-902), USP monograph isocratic MeOH/water 80:20 and community ACN/phosphate pH3 alternative, dilute-and-shoot + ACN-freeze-precipitation sample prep, 5-level USP RS calibration ($297/100 mg), USP <621> system suitability, identity (RT ±2% + DAD match ≥0.998) and potency (90–110%) acceptance, full troubleshooting table, and worked pass/under-dose/wrong-ester examples. ### DD2 — [`sop-oil-lle-gelclot-lal.md`](sop-oil-lle-gelclot-lal.md) (6,940 words) Aqueous liquid–liquid extraction + gel-clot LAL SOP for oil-based HRT endotoxin. Leads with the correct caveat: LLE is validated only for kinetic turbidimetric LAL (PPD Labs / Lester et al. *Am. Pharm. Rev.* 2019 on fractionated coconut oil); gel-clot extrapolation requires per-matrix revalidation with 2λ CSE spike recovery. Includes per-matrix behaviour for MCT, grapeseed, castor, cottonseed, sesame, benzyl benzoate/alcohol, and ethyl oleate; verified ACC part numbers (Pyrotell $595.35; corrected TB240 → TS050-10 for gel-clot reaction tubes); worked MVD math for testosterone cypionate (1,400× at λ=0.25) and estradiol valerate; mandatory PPC/NPC protocol; **250 °C / 30 min primary or 200 °C / 60 min Ph. Eur. depyrogenation (not 180 °C)**; lab notebook template; 16 specific failure modes. ### DD3 — [`peptide-testing-deep-dive.md`](peptide-testing-deep-dive.md) (5,861 words) The peptide-specific companion to Round 1's HRT-heavy corpus. Covers why peptides differ (MW 500–6000 Da, UV at 214 nm if no aromatics, sequence confirmation requires LC-MS/MS not HPLC-UV), detection strategy trade-offs (TFA noise at 214 nm vs formic acid for MS), MS/MS fragmentation basics, and a per-peptide detail table for semaglutide, tirzepatide, retatrutide, BPC-157, TB-500, CJC-1295 ±DAC, ipamorelin, GHK-Cu, melanotan-II, and tesamorelin — each with verified molecular formula/weight, aromatic residue content (BPC-157 has **zero** aromatics — key home-lab pitfall), recommended column chemistry, and Finnrick test counts. Finnrick aggregate: ~1-in-3 fails; ~8% endotoxin-positive on paid-tier peptides. ### DD4 — [`community-lab-profiles/`](community-lab-profiles/_index.md) (11 profiles + index) One-file-per-lab deep profiles with the same schema: status / location / principals / URL / instruments table / assays / pricing / submission / sample requirements / publication practice / reputation notes / structured claims / unknowns / sources. Labs covered: Janoshik, MZ Biolabs, Krause Analytical, BTLabs, Chromate, TrustPointe, Freedom Diagnostics, Trans Harm Reduction, Artemis Analytical (intentionally sparse), WEDINOS, DrugsData/Erowid. Critique items M12 (MZ Biolabs techniques page source), M13 (Janoshik staff/throughput attributed to Peptide Protocol Wiki), M14 (SARM/Package A prices flagged C3), M15 (THR GC-MS vs testing.trans.diy HPLC reconciled) all baked in. "Unknown" is a valid entry where primary sources don't document a fact. ### DD5 — [`annotated-chromatograms.md`](annotated-chromatograms.md) (6,218 words) Eight textual walkthroughs (no images; describe-what-you'd-see): passing EV (99.2% label claim, DAD match 0.9995), under-dosed same-identity vial (68.0% — "nothing wrong except the potency"), wrong-ester EEn-sold-as-EV (RT shift 8.2 → 6.5 min, DAD match 0.945), oxidation + plasticizer contamination, process-impurity profile at 0.1–0.3% area, semaglutide +16 Da oxidation on LC-MS, ghost peaks in blank injections, and column-aging signatures across 250 injections. Each scenario has the full "sample / method / what you'd see / interpretation / report text / action / citation" structure. ### DD6 — [`endotoxin-kit-comparison.md`](endotoxin-kit-comparison.md) (5,729 words) Side-by-side shootout of six kit families: ACC Pyrotell (gel-clot), Lonza Kinetic-QCL (kinetic chromogenic LAL), Lonza PyroGene (rFC), Charles River Endosafe-PTS+ (cartridge reader), Hyglos/Biosynth EndoZyme II (rFC), FUJIFILM Wako PYROSTAR ES-F/ES-J (dual-mode). Sensitivity ranges, reader requirements, regulatory matrix (USP <85>/<86>, EP 2.6.14/2.6.32, JP rFC status flagged C4/C5), cost scenario analysis at 1 / 50 / 500 tests/year, animal-free ethics. Critique items M1 (EP 2.6.32 effective 2021-01-01, not 2016), M8 (Pyrotell ≠ PYROSTAR sensitivities), M9 (USP <86> "alternative, not applicable to any monograph yet") all corrected. ### DD7 — [`used-hplc-buyers-guide.md`](used-hplc-buyers-guide.md) (7,185 words) Practical buyer's guide: Agilent 1050 → 1100 → 1200 → 1260 Infinity → 1260 Infinity II generation map with ChemStation B.04.03 / OpenLab CDS distinction; module-level failures with Agilent part numbers (5063-6589 pump seal, 5062-8562 AIV, 5190-0917 D2 lamp); pre-purchase inspection checklist with acceptance tests (caffeine injection at 273 nm, pressure-ripple, 6-rep CV); software license traps (ChemStation dongle, OpenLab transfer, Waters Empower workarounds); aftermarket parts ecosystem (Idex, Sciencix, Restek); refurb vendor comparison (Marshall Scientific, Conquer Scientific, LabX, eBay tiers); DIY refurb scope; April 2026 price table with the M6 DIS-49376 $10,500 → $24,500 drift explicitly surfaced; 4-tier decision framework. ### DD8 — [`aggregate-test-stats.md`](aggregate-test-stats.md) (5,409 words) Pooled community testing statistics with "lower bound of problems / upper bound of quality" framing front-loaded. Primary sources: testing.trans.diy (n=4, 100% pass, EEn-only), Trans Harm Reduction (n=16, 88% pass, 1 FAIL, 1 CONDITIONAL), Finnrick (n=6,813, 15 peptides, 204 vendors — with full computed grade distribution: A+B = 32.8%, D+F = 42.8% across 722 vendor-product pairs). Per-compound-class breakdown, per-vendor commentary strictly limited to public URL-backed findings, explicit acknowledgement of what these numbers do NOT measure (oil substitution, sterility failures, sequence-level errors undetected by UV-only testing). --- ## What this round changes about the Round 1 synthesis 1. **Silver-tier HPLC is no longer just a price point — it's a runnable method.** DD1's SOP converts "buy an Agilent 1100" into "buy these specific modules, run this method, pass these specific criteria." Combined with DD7's module-level failure guide, the path from eBay listing to first defensible EV potency result is concrete. 2. **Oil-LLE for home endotoxin testing is now documented with the correct caveats.** Round 1 implied gel-clot-after-LLE was a solved method. DD2 corrects the record: LLE was validated for kinetic turbidimetric only; gel-clot extrapolation carries the chemistry but not the quantitative behaviour. Per-matrix revalidation is mandatory. This makes the workflow harder to sell but safer to use. 3. **Peptide testing has its own operator handbook now.** BPC-157's zero aromatic residues — which make 280 nm detection useless — is the sort of trap that separates a lab that works from a lab that thinks it works. DD3 catalogs these trap-per-peptide. 4. **The community-lab ecosystem is documented at profile-depth.** Round 1 listed the labs; Round 2 gives you the instruments, pricing, submission flow, reputation notes, and unknowns for each of 11. Artemis's opacity is now documented as opacity, not hidden by aggregation. 5. **Aggregate stats are now cited, not assumed.** The "gray market is roughly 1-in-3 failure on peptides, small-n-unknown on HRT" narrative now has a computed Finnrick distribution behind it — with the caveats loud enough that it can't be weaponized into vendor libel. --- ## What Round 2 does **not** close Cross-reference [`../gaps.md`](../gaps.md) for the full list. Deep-dive-specific residuals: - **USP Estradiol Valerate monograph literal text** — still paywalled; DD1's method remains C3 reconstruction from secondary sources. - **Per-oil-matrix gel-clot validation for anything other than MCT** — no published source exists; DD2 is honest about extrapolation. - **MZ Biolabs founder identity** — about page 404s; techniques page names instruments but not people. - **Krause methodology Q&A content** — PDF is image-only, not OCR'd. - **THR's EU contract lab** — still unnamed. - **Time-series on Finnrick** — not exposed publicly; DD8 can't plot quarter-over-quarter. Each of these is flagged in-document and in `gaps.md`; none block the core deliverable. --- ## How to read this corpus - If you're **buying equipment**: DD7 → DD1 module list → Round 1 `../04-equipment-market/`. - If you're **setting up endotoxin testing**: DD6 kit comparison → DD2 oil SOP → Round 1 `../03-microbial-pyrogen/home-lab-endotoxin-feasibility.md`. - If you're **validating peptides**: DD3 → DD4 profiles of MZ Biolabs + Krause → Round 1 `../02-analytical-chemistry/lc-ms.md`. - If you're **picking a lab to send to**: DD4 profiles → DD8 stats → Round 1 `../10-cost-benefit/when-to-diy-vs-outsource.md`. - If you're **interpreting a chromatogram you got back**: DD5 scenarios → Round 1 `../08-interpretation-qc/`. - If you're **assessing market quality**: DD8 stats with the self-selection caveats. --- ## File map ``` research/deep-dive/ ├── index.md ← you are here ├── sop-estradiol-valerate-hplc.md (DD1) ├── sop-oil-lle-gelclot-lal.md (DD2) ├── peptide-testing-deep-dive.md (DD3) ├── community-lab-profiles/ (DD4) │ ├── _index.md │ ├── janoshik.md │ ├── mz-biolabs.md │ ├── krause-analytical.md │ ├── btlabs.md │ ├── chromate-labs.md │ ├── trustpointe.md │ ├── freedom-diagnostics.md │ ├── trans-harm-reduction.md │ ├── artemis-analytical.md │ ├── wedinos.md │ └── drugsdata-erowid.md ├── annotated-chromatograms.md (DD5) ├── endotoxin-kit-comparison.md (DD6) ├── used-hplc-buyers-guide.md (DD7) └── aggregate-test-stats.md (DD8) ```