Ref F1
Computational Service
Confidence Reported
Scoped After Feasibility Review
Service Overview
Diffusion and flow-based generative models build molecules directly in the three-dimensional context of a binding site, placing atoms where the pocket rewards them instead of proposing structures for docking to judge afterwards. We run structure-conditioned generation for full molecules, for fragment growing and linking, and for scaffold hopping under shape and pharmacophore constraints. Because generative output is cheap and uneven, the service is built around what happens after generation: filtering, physics-based confirmation and synthesis assessment.
Key Features & Advantages
Three-Dimensional Conditioning
Generation is conditioned on pocket geometry and chemistry, producing molecules whose shape and interactions are complementary by construction rather than by later selection.
Fragment Growing And Linking
Where fragment hits exist, models grow or link them while holding the crystallographic binding mode fixed, preserving the interaction that was experimentally observed.
Constraint-Driven Sampling
Pharmacophore requirements, substructure retention, property windows and excluded chemistry are imposed during generation, so output lands inside the design space you can actually use.
Physics Confirms Everything Delivered
All delivered molecules are re-docked, energy-checked and, for the top tier, simulated, because a generative model expresses a learned prior and not a binding measurement.
Where This Service Is Used
1Designing novel chemical matter for a well-characterised pocket
2Growing fragment hits into lead-like compounds while retaining the observed binding mode
3Linking two fragments bound in adjacent sub-pockets
4Scaffold hopping to escape an IP position or a scaffold-associated liability
5Expanding the accessible design space around an existing series
Methods and Techniques Applied
1Pocket preparation and encoding, with pharmacophoric and shape constraints derived from hot spots or known ligands
2Structure-conditioned diffusion or flow-based generation, sampling with varied noise schedules and seeds
3Validity, uniqueness, novelty and property filtering applied to the raw generated set
4Synthetic accessibility scoring and retrosynthetic assessment on the surviving candidates
5Re-docking, strain analysis and MM/GBSA rescoring for the final selection tier
6Clustering by scaffold with novelty comparison against known actives and published chemotypes
What We Need From You
1Target structure with a clearly defined site and, ideally, a reference ligand pose
2Known actives, fragments or crystallographic poses to condition and anchor the generation
3Design constraints: property windows, substructures to keep, chemistry to avoid, novelty requirements
4The number of compounds you can realistically synthesise, which determines selection size
What You Receive
1Generated molecule set with SMILES, three-dimensional poses and predicted property profiles
2Synthetic accessibility assessment with proposed routes for the recommended compounds
3Re-docking and rescoring results confirming each delivered design
4Novelty and diversity analysis against known chemical matter
5Generation report covering model, constraints, sampling and attrition at each filter
What We Compute and Report
1Generated structures with poses in the binding site and per-design scoring
2Validity, uniqueness and novelty statistics across the generated set
3Property profile and developability flags for every delivered molecule
4Synthetic accessibility with route assessment for the recommended tier
5Confirmation scores from independent re-docking and energetic rescoring
Questions We Are Asked Before Scoping
How is this different from de novo small molecule design?
▸It is the same family of methods viewed from the platform side. This page describes the generative engine and its modes; the design service describes a full programme built around it. Projects usually buy the programme.
Are generated molecules patentable?
▸Novelty of structure is a necessary but not sufficient condition, and patentability is a legal judgement. We provide structural novelty analysis; your attorneys make the determination.
Do generative models produce better molecules than screening?
▸They produce different ones — novel and pocket-shaped, but unmeasured. Where a large library screen is viable, running both and comparing is usually the stronger strategy.
Interpretation and Limitations
Computational predictions prioritise hypotheses; they do not independently establish binding, biological activity, therapeutic efficacy, safety or clinical validity. Performance depends on input quality, target class, the structural evidence available and the applicability domain of the selected methods. Scores used for ranking are comparative devices and should not be read as measured quantities. All deliverables are for research use only and are not intended for diagnostic or therapeutic use.
Is This the Right Service for You?
Choose this when
A pocket is well characterised and novel chemistry is needed inside it
Fragments need growing or linking while preserving an observed binding mode
A scaffold hop is required under explicit shape and pharmacophore constraints
Consider another service when
The scaffold is fixed and several properties must improve at once
Reinforcement Learning for Molecule Optimization Purchasable compounds would answer the question sooner
High-Throughput Virtual Screening What Goes Wrong, and How We Handle It
These are the failure modes that produce confident-looking results which do not survive the next experiment. Each one is designed against rather than discovered afterwards.
Vast generated sets that are mostly unusable
Generation is the cheap step; the delivered set is what survives validity, property, accessibility, novelty and re-docking filters, with attrition reported at each stage.
Three-dimensional proposals that ignore ring strain and geometry
Strain analysis and energy checks are run on every delivered pose, not only scoring.
Novel-looking molecules that are minor variants of known actives
Similarity distance to known actives and published series is computed and delivered per design.
A Worked Example
Scaffold hop out of a patented series
The situation
The pharmacophore was well understood but the core was covered by a competitor filing.
What the work involved
Pharmacophoric constraints were extracted from the existing series, generation was conditioned on the pocket with the core excluded, and output was filtered for accessibility, properties and similarity to the patented chemotype.
What came back
Sixty designs across five new cores, each preserving the three interactions responsible for potency, with routes sketched for the twelve most accessible.
Illustrative of the shape this work takes. Details are generalised, no client or target is identified, and scope and outcome vary with the system in question.
What Published Work Reports
The results below were published by independent groups and are cited to show what these computational approaches have achieved in the literature. They are not ICDMO project results, and outcomes vary substantially between targets, datasets and programmes. We do not promise comparable results for any particular project.
A generative-AI-designed inhibitor reported positive Phase 2a lung function data
A programme in which both the target and the molecule were identified using AI methods reached a randomised Phase 2a trial in idiopathic pulmonary fibrosis. Across 71 patients treated for 12 weeks, the 60 mg once-daily arm showed a mean forced vital capacity change of +98.4 mL against −20.3 mL in the placebo arm, with a manageable safety profile. It remains an early-phase result in a small trial, but it is the first published clinical proof of concept for an AI-discovered target and AI-designed compound.
Nature Medicine, 2025 ↗Service Workflow
1
Scientific Consultation
We discuss the objective, the decision it supports and the data you hold. Nothing is quoted before we understand what the result has to answer.
2
Feasibility Review
We check that the data, target class and available structural evidence can support the proposed approach, and say plainly where they cannot.
3
Technical Proposal
Scope, methods, deliverables, timeline and price are confirmed in writing, together with the limitations we expect to report.
4
Computation and QC
The analysis runs on pinned tool versions in a containerised environment, with quality checkpoints and method comparison where that is informative.
5
Report and Follow-Up
You receive the ranked results, the documented methods, the limitations and a proposed validation plan, plus support in interpreting them.
Service Process
1
Online Consultation
›
2
Solution Matching
›
3
Service Contract
›
4
Project Execution
›
5
Data Delivery
Specifications & Turnaround
Note: ICDMO delivers computational (in silico) services. Results prioritise hypotheses for testing; they do not by themselves establish binding, biological activity, efficacy or safety. All deliverables are for research use only and are not intended for diagnostic or therapeutic use. Downstream experimental work is scoped separately and executed through qualified partners.
Get a Custom Quote
Our scientific team responds within 24 hours with a detailed technical proposal and pricing tailored to your project.
Contact Us Online ConsultationStandard Deliverables
✓Sequence-verified final product
✓Full QC documentation
✓Certificate of Analysis (CoA)
✓Electronic data files
✓Project completion report
Why ICDMO?
Computational CDMO Focus
Reproducible, Versioned Pipelines
Confidence & Limitations Reported
Experiment-Aware Recommendations
Response Within One Business Day