HomePlatformEnterpriseThe ScienceAboutCompanyInsightsPublicationsPrivacyTerms
DeepCeutix - AI Drug Design PlatformDeepCeutix - AI Drug Design Platform
Platform
Resources
Enterprise
Company
  • 01Platform
    OverviewResearch agentsBiologics agentsSafety agents
  • 02Resources
    The ScienceInsightsPublications
  • 03Enterprise
  • 04Company
    AboutPress KitContact
DeepCeutix - AI Drug Design PlatformDeepCeutix - AI Drug Design Platform

Autonomous Pharmaceutical Intelligence.
London, UK

Try the playground

Platform

  • Platform
  • Research agents
  • Biologics agents
  • Safety agents
  • Enterprise

Resources

  • The Science
  • Strategic briefings
  • Publications

Company

  • About
  • Contact
  • Press Kit

Trust

  • Trust Centre
  • Privacy
  • Terms
All Systems Operational
© 2026 DeepCeutix Ltd. // Engineered in London
© 2026 NVIDIA, the NVIDIA logo are trademarks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries.

Safety suite

Six safety
specialists.

ERA, extractables, leachables, nitrosamines, OEL, and PDE. Each agent reads from named regulatory texts and returns the deliverable filed in the dossier, not a literature summary. Anchored to ICH M7(R2), ICH Q3E, EMA HBEL, EMA Appendix 1, USP <1663>/<1664>, PQRI PODP, and ECETOC TR 101.

Fig 01 · Anchored to regulation

Six agents.
Six named anchors.

Each safety specialist files against a named regulatory text. ERA against the EMA Rev. 1 Environmental Risk Assessment guideline. Extractables and Leachables against ICH Q3E, USP <1663>/<1664>, and PQRI PODP. Nitrosamines against ICH M7(R2) and EMA Appendix 1. OEL against ECETOC TR 101. PDE against the EMA HBEL guideline.

DeepCera-agentERAAGENTEMA Rev. 1EMEA/CHMP/SWP/4447/00CTD 1.6.1extractables-agentEXTRACTABLESAGENTICH Q3EUSP <1663> · PQRI PODPleachables-agentLEACHABLESAGENTICH Q3EUSP <1664> · PQRI PODPnitrosamines-agentNITROSAMINESAGENTICH M7(R2)EMA App. 1 · ng/day AIoel-agentOELAGENTECETOC 101ISPE Risk-MaPP · OEB 1-5pde-agentPDEAGENTEMA HBELCHMP/CVMP/SWP/169430

Click any slab to jump to that agent. Every output is filed in the framework's prescribed structure, with a Klimisch-tiered citation annex and an unsigned QP sign-off block.

The shared five-phase methodology

Every safety dossier
runs five phases.

01

Research

Substance identity from PubChem and ECHA. Primary toxicology study harvest from REACH IUCLID. Threshold lookup across EFSA OpenFoodTox, JECFA, EPA IRIS, EPA CompTox, EMA Appendix 1. Occupational limits from NIOSH, OSHA, GESTIS.

02

Read-across

When the target lacks direct data, DeepC selects a structurally similar analog with characterised data. Toxicophore parity is verified by running PAINS / structural-alert analysis on both target and analog. F6 = 10 uncertainty factor by default.

03

Threshold derivation

Every safety factor is shown as an explicit multiplication, never a single opaque number. The QP can re-derive the result from the arithmetic shown.

04

Citation hygiene

Every primary toxicity-study citation carries a Klimisch reliability tier (Ks 1-4). Each cited study gets a 2-4 sentence narrative covering species + strain, doses tested, GLP status, vehicle / route / duration, and the source-verbatim conclusion.

05

Presentation

The full markdown report renders to a regulatory PDF in the framework’s prescribed structure: cover page, numbered references, Klimisch annex, unsigned QP sign-off block.

Per agent

Each specialist
in detail.

Click any anchor above to jump to its agent. Or scroll the six sections below.

01Safety

ERA Agent

EMA EMEA/CHMP/SWP/4447/00 Rev. 1In force 1 September 2024. Files in CTD Module 1.6.1

Environmental Risk Assessment for human medicinal products.

  • →Phase I Q1-Q7 decision tree and PEC_SW against the 0.01 µg/L action limit.
  • →Phase II Tier A: OECD 201 / 211 / 210 triplet + STP respiration (209).
  • →Tier B refinement: F_PEN, F_EXCRETA, SimpleTreat, mesocosm, SSDs.
  • →PBT/vPvB per REACH Annex XIII, independent of RQ outcome.

“Run an ERA on lacosamide 10 mg/mL solution for infusion.”

Example prompt

Per the EMA ERA Guideline (EMEA/CHMP/SWP/4447/00 Rev. 1, in force 1 September 2024). Phase I PEC/PNEC ratio screen and Phase II refinement workflow. Pulls aquatic-toxicity studies from ECHA IUCLID and physico-chem from PubChem. Phase II Tier A runs the OECD test guideline triplet (algae 201, Daphnia 211, fish FELS 210), STP respiration (209), sediment (218/219/225/233), and PNEC derivation per compartment. PBT/vPvB assessment per REACH Annex XIII runs independent of the risk-quotient outcome.

Returns: A 30-60 page ERA dossier, filed in CTD Module 1.6.1.

Cited studiesKlimisch reliability
  • Ks 1

    ECHA IUCLID Dossier 100.014.002

    EC50 (Daphnia magna, 48 h) · ECHA REACH

    GLP-compliant OECD 202 acute immobilisation study. EC50 = 4.2 mg/L (95% CI 3.6-4.9). Used to derive PNEC_water with assessment factor of 1000.

  • Ks 1

    ECHA IUCLID Dossier 100.014.002

    NOEC (algae, 72 h) · ECHA REACH

    GLP-compliant OECD 201 growth-inhibition study on Pseudokirchneriella subcapitata. NOEC = 0.3 mg/L based on growth-rate endpoint.

  • Ks 2

    EMA EPAR Scientific Discussion, 2024

    PEC_SW (refined) · EMA

    Refined PEC_SW after applying F_PEN = 0.0017 from epidemiological prevalence data. PEC_SW / PNEC_SW < 1 in all assessed compartments.

Fig. Klimisch-tiered citation list returned by the ERA Agent (illustrative).

02Safety

Extractables Agent

ICH Q3E · USP <1663> · PQRI PODPPQRI PODP AET 1.5 µg/day · 5 µg/day for sensitisers

Container/closure stress-extraction qualification.

  • →Stress-extraction context (solvent, temperature, duration, SA:V, detection) captured.
  • →Per-substance worst-case daily exposure from extractable mass × units/day.
  • →Route correction oral PoD → IV via ICH Q3C(R9) Appendix 3 α₁/α₂.
  • →Read-across with named class analogs (e.g. CAS 36306-87-3, butyl rubber).

“Qualify the extractables identified in our pre-filled syringe stopper study against an IV dose.”

Example prompt

Per ICH Q3E + USP <1663> + PQRI PODP. Component categorization, screening study workup, identification, and toxicological assessment for the proposed container/closure or single-use system. Stress-extraction context (solvent system, temperature, duration, surface-area:volume ratio, detection method) captured in the report introduction. Read-across via structural similarity uses named class analogs (CAS 36306-87-3 for bromobutyl/chlorobutyl rubber, for example) when direct toxicology data are missing.

Returns: An extractables report with the analytical workup and a tier-annotated citation list.

RETURNSExtractablesreportAETper-substancequalification
Fig. Extractables Agent deliverable.
03Safety

Leachables Agent

ICH Q3E · USP <1664> · PQRI PODPICH M7(R2) less-than-lifetime brackets · EMA Appendix 1

In-use migration safety thresholds.

  • →TDI = concentration × maximum daily dose volume.
  • →ICH M7(R2) LTL brackets: ≤1mo → 120 / 1-12mo → 20 / 1-10y → 10 / >10y → 1.5 µg/day TTC.
  • →PQRI PODP 5 µg/day floor for sensitisers in parenteral and ophthalmic products.
  • →EMA Appendix 1 lookup when leachable falls in the cohort-of-concern.

“Assess the leachable risk for formaldehyde detected in our 12-month accelerated stability.”

Example prompt

Per ICH Q3E + USP <1664> + PQRI PODP. End-of-shelf-life leachables study design, AET calculation, and toxicological qualification. Cross-references the matching extractables case in the file. ICH M7(R2) less-than-lifetime brackets quoted explicitly: ≤1 month → 120 µg/day, 1-12 months → 20 µg/day, 1-10 years → 10 µg/day, >10 years → 1.5 µg/day TTC. PQRI PODP threshold of 5 µg/day for sensitisers/irritants in parenteral and ophthalmic products.

Returns: A leachables assessment report with AET derivation and per-leachable risk tables.

Per ICH Q3E + USP <1664> + PQRI PODPCase file lea-7421-A
LEACHABLES ASSESSMENT

Lacosamide 10 mg/mL solution for infusion

SponsorAcme Pharmaceuticals plc
Prepared2026-04-22
Sign-off (left blank for Qualified Person)
Name
Signed
Date

Fig. Leachables Assessment cover page returned by the Leachables Agent (illustrative).

04Safety

Nitrosamines Agent

ICH M7(R2) · EMA Appendix 1 · FDA NDSRI GuidanceEMA Appendix 2 CPCA Category 1-5 for novel NDSRIs

Acceptable Intake derivation in ng/day.

  • →Three AI paths: TD50-based (AI = TD50 ÷ 50,000 × 50 kg), CPCA Cat 1-5, or EMA App. 1 direct lookup.
  • →LTL adjustment per ICH M7(R2) §7.3 Table 2 for sub-lifetime treatment.
  • →Specification limit in ppm with method LOQ ≤ 30% of spec.
  • →Root-cause mechanism: amine + nitrite, photo-degradation, packaging migration, nitrite-cured rubber.

“We have a novel NDSRI from our secondary amine API + nitrite in excipients. Run a CPCA categorization.”

Example prompt

Per ICH M7(R2) + EMA Appendix 1 + EMA Appendix 2 (CPCA categorization for novel structures) + FDA Guidance on Control of Nitrosamine Impurities (Aug 2023). Output is in ng/day. Three AI derivation paths: TD50-based (AI = TD50 ÷ 50,000 × 50 kg), CPCA Category 1-5 for novel NDSRIs with no carcinogenicity data, or direct EMA Appendix 1 lookup. Less-than-lifetime adjustment per ICH M7(R2) §7.3 Table 2.

Returns: A nitrosamines risk assessment with CPCA category, AI in ng/day, and a control-strategy memo.

Cited studiesKlimisch reliability
  • Ks 1

    ECHA IUCLID Dossier 100.005.235

    NOAEL (rat, 90-day oral) · ECHA REACH

    GLP-compliant 90-day oral gavage study in Sprague-Dawley rats (5/sex/dose) at 0, 50, 150, 500 mg/kg/day. NOAEL of 150 mg/kg/day established; LOAEL of 500 mg/kg/day driven by hepatic centrilobular hypertrophy.

  • Ks 2

    EFSA peer-reviewed re-evaluation (2023)

    NOAEL (rat, 90-day oral) · EFSA

    Authoritative re-evaluation citing the underlying ECHA dossier. Concurs with NOAEL of 150 mg/kg/day for the same critical effect.

  • Ks 2

    Mortelmans et al. PMID 3781131

    NOAEL (rat, 90-day oral) · PubMed

    Peer-reviewed primary publication of the same 90-day oral toxicity study program. Methodology documented in detail; conclusions match the ECHA dossier.

  • Ks 3

    NTP TR-403 (secondary citation, trade publication 2019)

    NOAEL (rat, 90-day oral) · Industry whitepaper

    Reports a NOAEL of 50 mg/kg/day citing NTP TR-403. The cited NTP study uses a different strain (F344, not Sprague-Dawley) and a different route (drinking water, not gavage).

    Rejected by Reality Anchor: Strain and route mismatch; superseded by Ks 1 ECHA dossier.

Fig. Klimisch citations with one entry rejected by the Reality Anchor.

05Safety

OEL Agent

ECETOC TR 101 · Sussman/Naumann/Pfister 2016ISPE Risk-MaPP · ASTM E2476 · OEB 1-5

Occupational Exposure Limit and OEB containment band.

  • →Worker receptor model: 70 kg, 10 m³/shift, worker-relevant critical effects.
  • →Composite UF per-component (animal-human, intra-human, duration, LOAEL→NOAEL, database, MF).
  • →OEB 1-5 tier with ISPE Risk-MaPP containment + RPE APF.
  • →Hard floors: 0.15 µg/m³ for genotoxics (M7 TTC) · < 0.01 µg/m³ for respiratory sensitisers.

“Set an OEL and OEB for our highly potent oncology API for tablet manufacturing operations.”

Example prompt

Per Sussman/Naumann/Pfister (Tox Sci 2016) and ECETOC TR 101. The OEL drives the OEB band, which drives engineering controls (open handling vs. local exhaust ventilation vs. isolator vs. glovebox) per ISPE Risk-MaPP and ASTM E2476. Output in µg/m³ airborne 8-hour TWA. Worker-specific receptor model: 70 kg body weight, 10 m³/shift breathing volume. Special handling for genotoxic APIs (M7 TTC fallback: 1.5 µg/day systemic → 0.15 µg/m³ airborne) and respiratory sensitisers (hard floor < 0.01 µg/m³).

Returns: An OEL/OEB report with POD selection rationale, AF table, OEB band assignment, and ISPE Risk-MaPP containment recommendation.

RETURNSOELOEB bandµg/m³ TWAcontainment tier · RPE
Fig. OEL Agent deliverable.
06Safety

PDE Agent

EMA HBEL (EMA/CHMP/CVMP/SWP/169430/2012)MACO worked examples · ICH Q3D parallel

Permitted Daily Exposure for shared-facility cleaning.

  • →Core derivation: PDE = (NOAEL × 50 kg) / (F1 × F2 × F3 × F4 × F5), per-factor justified.
  • →Critical-effect selection with full tabulation across studies, species, durations, endpoints.
  • →Route extrapolation with bioavailability-corrected PDE when NOAEL route ≠ target.
  • →Special handling: sensitisers (dedicated facility), macromolecules (often §5.3 exempt), genotoxicants (TTC fallback).

“Derive a PDE for lacosamide for a shared oral-solid-dose facility.”

Example prompt

Per the EMA HBEL Guideline (EMA/CHMP/CVMP/SWP/169430/2012). Core derivation: PDE = (NOAEL × 50 kg) / (F1 × F2 × F3 × F4 × F5), with per-factor justification of species, inter-individual variability, study duration, severity, and NOAEL/LOAEL adjustment. Critical effect selection with full tabulation. MACO worked examples in mg per equipment area and ppm in next-product batch. Special handling for highly sensitising APIs (dedicated facility per GMP §3.6), therapeutic macromolecules (often exempt under EMA §5.3), and genotoxicants without threshold (TTC fallback per EMEA/CHMP/QWP/251344/2006).

Returns: A PDE monograph in EMA HBEL format with MACO worked examples.

RETURNSPDEmonographHBEL formatMACO worked
Fig. PDE Agent deliverable.

Research suite

Seven research specialists

Formulation, optimization, research, FTO, patent drafting, VCM process design, and analytics.

Biologics suite

Three biologics specialists

Biologics Research, FTO, and Formulation & CMC, anchored to ICH Q5A-E, Q6B, Q13, Q14, FDA Purple Book, EMA biosimilar register, SAbDab, and PLAbDab.

Architecture

How the platform routes

Reality Anchor, agent routing, foundation, regulatory frameworks, and data sources: the unified platform overview.

Book a demo.

Tell us what you are working on (molecule, indication, stage, open question).

A formulation lead will be in touch within a business day.

By submitting, you agree to our Privacy Policy.


info@deepceutix.com|London, UK|LinkedIn