ADMET & Toxicity Research Screening
Run bounded dry-lab structural screening before deciding which evidence warrants deeper experimental review.
Who this is for
Drug-discovery program leads, computational toxicology scientists and technical-diligence reviewers screening early research candidates.
The problem you already have
A candidate can look acceptable in one screen and still carry a safety contradiction elsewhere
Early discovery teams often review structural plausibility, ADMET signals, off-target similarity and toxicity flags in separate tools. Part 34's safety-ledger approach treats those stages as one evidence chain, so a candidate does not receive a stronger screening status merely because one isolated score looks favorable.
“Drug-like” is not the same as “safe enough to advance”
The documented workflow keeps specific rule-gate evidence such as hERG liability, blood-brain-barrier behavior and CYP3A4-related risk separate from the broader structural proxy. That matters because a candidate can look structurally interesting while still carrying a specific downstream safety flag that should stop or downgrade the review.
A dry-lab proxy must never be mistaken for validated assay or clinical evidence
The supported capability is a research-only computational screen. It does not establish efficacy, validated binding, synthesis readiness, dosing, treatment suitability or clinical safety. Experimental wet-lab validation remains mandatory before a candidate advances beyond computational review.
Real-World Operational Scenario: The Candidate That Looked Clean Until the Safety Ledger Was Read as a Whole
Scenario: A lead compound has a promising structural score, but one safety stage conflicts with the rest of the package.
A discovery team has dozens of candidates and limited assay capacity. One candidate ranks well on structural evidence and looks attractive enough to move forward, but the broader evidence package also contains an off-target similarity signal and a rule-gate safety flag.
WITHOUT this check: The promising structural result dominates the review, and the candidate consumes deeper assay or CRO budget before the contradictory safety evidence is reconciled.
WITH this check: The candidate is evaluated through the deterministic multi-stage safety ledger. Structural proxy evidence, Morgan-fingerprint off-target similarity and named safety rule gates are preserved together, and contradictory evidence downgrades or holds the candidate for human research review instead of being averaged away.
What BivectorAI Provides
Deterministic Multi-Stage Drug Safety Ledger
Connects structural screening, ADMET proxy evidence, off-target similarity and contradiction handling into one reviewable evidence chain instead of a collection of unrelated scores.
Named Safety Rule Gates
Preserves specific rule-gate evidence including hERG, BBB and CYP3A4-related screening flags so a broad favorable score cannot silently override a concrete safety concern.
Morgan-Fingerprint Off-Target Similarity
Surfaces similarity to toxic or off-target reference structures as a separate evidence item, keeping chemical-neighborhood risk visible during candidate triage.
Research-Only Evidence Boundary
Every result remains computational screening evidence for human research review. It does not become a clinical recommendation, efficacy claim, validated binder result or substitute for experimental validation.
Prove the Gate Before Broader Integration
Do not buy the screening claim. Back-test it against candidates whose downstream disposition is already known.
Start with a fixed historical candidate set and freeze the supplied evidence scope before revealing which candidates later failed or advanced. The evaluation is useful only if the ledger consistently surfaces contradictions and weak candidates early enough to improve human triage without promoting unsupported claims.
What you provide
A bounded, appropriately redacted candidate representation together with the candidate-level structural or physicochemical descriptors and research-context fields already available for the agreed dry-lab screen. Patient data, treatment context and clinical decision authority are not required.
What BivectorAI actually hands you
Candidate safety-ledger result
A bounded research-screening status tying the candidate to the evidence stages actually evaluated.
Named ADMET / toxicity flags
Research-only rule-gate findings, including applicable hERG, BBB and CYP3A4-related evidence where supported by the supplied candidate representation.
Off-target similarity evidence
Morgan-fingerprint similarity evidence retained separately from structural plausibility so contradictory chemical-neighborhood risk remains visible.
Validation boundary
Explicit statements of what remains unvalidated and what requires downstream experimental review before any stronger scientific or clinical conclusion.
What we will never ask you for
This is a bounded dry-lab research screen, so patient information, treatment authority and claims of validated assay or candidate performance stay outside the engagement.
- Patient data.
- Dosing or treatment authority.
- A claim of validated assay performance.
- A claim that a screened candidate is efficacy-, binder- or synthesis-validated.
How it actually runs
Start with a bounded evaluation
Fixed scope / quoted per engagement. Inputs, outputs, evidence boundary and acceptance criteria are agreed before work begins.
We do not open with an annual license. If the bounded ADMET and toxicity screen does not establish that the research evidence is useful for the agreed candidate-review question, there is no reason to move into a broader commercial commitment.
Current capability is a module-only dry-lab research proxy. It is not a validated ADMET/toxicity assay, clinical recommendation, drug-efficacy determination, validated binder assessment or synthesis-ready certification.