Computational pharmacology

Drug intelligence, traced back to primary evidence.

Every promising molecule begins as a question: will it bind, will it last, will it be safe. Docking, molecular dynamics, ADMET, pharmacogenomic and pharmacovigilance analysis — written up so your team can hand it to a regulator, an investment committee or a medicinal chemist.

Physician-reviewed before delivery — every page is read end to end by a person who understands the pharmacology, not generated and shipped unread.

Curated entities
17,721Small molecules, biologics and gene therapies
Cross-domain assets
1,114Compounds active in two or more therapeutic domains
Therapeutic domains
8Cardiometabolic through longevity
Literature grounding
PubMedLive NCBI retrieval, citations resolvable

What we analyse

Four lines of evidence, one document.

Most vendors sell a database query. The value is the interpretation — what the signal means for your asset, your protocol and your next decision, read and signed off by a person.

Pharmacovigilance & safety

Adverse effects led by the peer-reviewed literature and the label, with spontaneous reports read as a signal against the labelled indication.

  • Effects named in the literature, each with citable PMIDs
  • Boxed warnings and adverse-reaction sections from the label
  • Class-effect comparison against analogues
  • Withdrawn-drug precedent analysis

Pharmacogenomics

Metabolic liability mapped to the enzymes that actually clear the compound, including star-allele consequences.

  • CYP450 and transporter pathway assignment
  • Star-allele phenotype consequences
  • Prodrug bioactivation dependency
  • Boxed-warning and label precedent

Structure, docking & dynamics

Structure-based work on a named target, with parameters stated so a run can be checked or repeated.

  • AutoDock Vina docking with justified box placement
  • GNINA convolutional rescoring and decoy controls
  • Molecular dynamics and MM/GBSA relative ranking
  • Pose interpretation, not just a score

ADMET & liability modelling

Purpose-built models for the off-target risk that would stop a programme, with the uncertainty reported rather than hidden.

  • Models trained on curated ChEMBL bioactivity
  • Cross-validated, with an applicability-domain check
  • Cardiac, barrier, efflux, mutagenicity and hepatotoxicity panels
  • Analogue and bioisostere design around a retained pharmacophore

Why this matters now

What computation changes, stage by stage.

Drug discovery has not stopped being slow and expensive. What has changed is where the expensive questions get asked: a liability that used to surface in a preclinical study can now be raised — and often ruled out — before a compound is made.

  1. 01

    Discovery

    Target selection

    • Structure retrieval from the PDB
    • Pathway and mechanism review
    • Live literature mining

    Asks: is this target plausible, and what is already known against it?

  2. 02

    Hit finding

    Virtual screening

    • Fingerprint similarity across curated chemical space
    • Docking into a justified pocket
    • Deep-learned pose rescoring

    Replaces: screening a plate to find out a series was never going to fit.

  3. 03

    Lead optimisation

    Design and liability

    • Analogue and bioisostere enumeration
    • Machine-learned off-target models
    • Molecular dynamics and binding-energy ranking

    Asks: which of a thousand analogues is worth a chemist’s week?

  4. 04

    Preclinical

    ADMET and genotype

    • Cardiac, barrier, efflux and hepatic panels
    • CYP and transporter pathway assignment
    • Star-allele consequence mapping

    Surfaces: the metabolic surprise that would otherwise appear in Phase 1.

  5. 05

    Clinical and beyond

    Stratify and monitor

    • Genotype-guided inclusion criteria
    • Reported adverse-event signal mining
    • Class and withdrawn-drug precedent

    Asks: who should be in this trial, and what has the class already done to people?

Earlier, not instead

Computation moves a question forward in time. A patch-clamp assay still settles cardiac risk — but it is cheap once you know which two compounds to run it on.

Negative results, early

The most valuable output is often a ruled-out hypothesis. A model that cannot separate your analogues is told to say so rather than produce a ranking.

Traceable, not oracular

Every figure resolves to a PDB code, a PMID, a database version or a stated parameter. Nothing rests on a number whose origin cannot be shown.

How it works

Free where a machine can answer.

The platform and the first look cost nothing, and they are not a trial. What is charged for is the part a machine cannot do.

Free, for everyone

Open to academics, scientists, industry teams and investors. No account, no card, no expiry.

  • The platform — about 17,721 molecules, searchable, with six ways to search
  • A one-page technical snapshot on any compound or ClinicalTrials.gov number
  • A short technical sanity-check for investors, on one pipeline asset

Where human work starts

Analysis that requires a trained reading: interpretation, design, and a written argument you can act on.

  • Pharmacogenomic panel evaluation and protocol-level recommendations
  • Technical due diligence memos ahead of an investment decision
  • Docking, dynamics, ADMET and analogue design on a named target

Scope, depth and turnaround differ too much between engagements for a list price to mean anything. Describe the compound and the decision it has to inform, and you will get a personalised offer.

No cost, no obligation

Send a compound. Get a technical snapshot back.

No account, no card. Tell us the compound and the question; the first thing back is a one-page technical snapshot at no charge.

Book a consultation

Prefer email? Write to . Either way the request reaches the same inbox.

Tell us the compound and the question.

Scope and turnaround come back within two working days, with a personalised offer where the work needs a person. If the question is outside what the method can honestly answer, we will say so rather than sell you a report.