Ref C1
Computational Service
Confidence Reported
Scoped After Feasibility Review
Service Overview
De novo design generates molecules that do not exist in any catalogue, shaped to a specific binding site. We use structure-conditioned generative models to propose chemistry that fits the pocket, then filter hard: synthetic accessibility, drug-likeness, predicted ADMET, novelty against published and patented matter, and re-docking of every survivor. Generation is the cheap part of this service; the value lies in the filtering and in a candidate list a chemist is willing to make.
Key Features & Advantages
Pocket-Conditioned Generation
Molecules are generated in the context of the three-dimensional site rather than sampled from a chemical distribution and docked afterwards, which improves shape and interaction complementarity.
Synthesis Realism Enforced
Every proposal is scored for synthetic accessibility and checked by retrosynthetic analysis, with a route sketch for the final set so proposals are not merely elegant on screen.
Multi-Objective From The Start
Potency, physicochemical properties, ADMET flags and novelty are optimised jointly, avoiding the familiar outcome of a high-scoring binder that fails on solubility or metabolism.
Novelty Checked Explicitly
Generated scaffolds are compared against known actives and patent-adjacent chemotypes so you learn which proposals are genuinely new and which merely rediscover the literature.
Where This Service Is Used
1Generating chemical starting points when virtual screening of existing libraries has failed
2Designing around a competitor patent while keeping the interactions that matter
3Growing, linking or merging fragment hits into lead-like molecules
4Scaffold hopping away from a series with a fixed liability such as hERG or metabolic instability
5Exploring a newly discovered allosteric or cryptic pocket with no known chemical matter
Methods and Techniques Applied
1Binding site preparation and definition of pharmacophoric constraints from hot spots or known ligands
2Structure-conditioned generative sampling, typically producing tens of thousands of raw candidates
3Filtering for validity, drug-likeness, property ranges, PAINS, reactive groups and predicted ADMET liabilities
4Synthetic accessibility scoring and retrosynthetic route assessment for the surviving set
5Re-docking and, for the final tier, MM/GBSA rescoring and short MD stability checks
6Scaffold clustering, novelty and freedom-to-operate style similarity analysis, and final ranked selection
What We Need From You
1Target structure with the site to be designed against, ideally with a reference ligand pose
2Known actives and the SAR behind them, which strongly improves the relevance of what is generated
3Property and liability constraints: molecular weight and logP ranges, chemistry to avoid, required selectivity
4Synthesis capability available to you, since it determines how ambitious the accepted routes can be
What You Receive
1Ranked set of designed molecules with SMILES, docked poses and predicted property profiles
2Retrosynthetic assessment with a proposed route for the recommended compounds
3Novelty analysis against known actives, published series and patent-adjacent chemotypes
4Interaction analysis explaining what each design is intended to achieve in the pocket
5Generation and filtering report with model choices, thresholds and the attrition at each stage
What We Compute and Report
1Generated structures with three-dimensional poses in the target pocket
2Predicted property profile per design: lipophilicity, solubility, polar surface area, drug-likeness flags
3Synthetic accessibility score with a retrosynthetic route assessment for recommended designs
4Novelty and similarity distance to known actives and published series
5Re-docking score and interaction analysis confirming every delivered design
Questions We Are Asked Before Scoping
Will the designs be synthesisable?
▸The delivered set is filtered and route-assessed for that reason, but accessibility scoring is imperfect and a chemist may still reject some proposals. We would rather deliver fewer, makeable compounds than a large list with a low make rate.
How potent will generated molecules be?
▸Unknown until tested. Generative models optimise predicted complementarity, which is a weak proxy for affinity. Typical programmes expect micromolar starting points to be refined, not nanomolar leads on the first cycle.
Who owns the designs?
▸You do. Deliverables, including generated structures, belong to the client under our standard agreement. We do not reuse client-specific designs for other projects.
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
Screening has not produced usable chemical matter for a well-characterised pocket
Fragment hits exist and need growing while preserving a crystallographic binding mode
A series must escape a patent position without losing the interactions that drive potency
Consider another service when
Purchasable compounds would answer the question faster
High-Throughput Virtual Screening A scaffold is fixed and the task is multi-property optimisation
Reinforcement Learning for Molecule Optimization 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.
Elegant structures that no chemist will make
Accessibility scoring plus retrosynthetic assessment is applied before delivery, and the recommended tier ships with a route sketch.
High predicted complementarity paired with hopeless physical properties
Property and liability filters are applied during generation rather than as a post hoc sort.
Generated molecules that merely rediscover the published series
Novelty distance against known actives and patent-adjacent chemotypes is computed and reported per design.
A Worked Example
Growing two fragments into a single lead-like molecule
The situation
Crystallography had placed two weak fragments in adjacent sub-pockets, separated by a narrow channel.
What the work involved
Both crystallographic poses were held fixed while linker chemistry was generated into the channel, then filtered for accessibility, strain and properties, and every survivor was re-docked and simulated briefly.
What came back
Forty designs ranked by predicted fit and makeability, with the six shortest synthetic routes highlighted and the vectors left open for later optimisation identified.
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.
AI-designed molecules have now produced clinical-stage compounds
A compound designed by generative methods against an AI-nominated target completed a randomised Phase 2a trial in idiopathic pulmonary fibrosis, reporting a mean forced vital capacity change of +98.4 mL in the 60 mg once-daily arm against −20.3 mL for placebo over 12 weeks in 71 patients. The result shows that generative design can produce molecules that survive to clinical testing; it does not show that any individual generated series will.
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