Ref A3
Computational Service
Confidence Reported
Scoped After Feasibility Review
Service Overview
Structure prediction has become reliable enough to carry design decisions, but only when the prediction is delivered with its confidence properly interpreted. We predict monomers, multimers and complexes — including protein with small molecule, peptide, nucleic acid or ion — using current deep-learning systems, then evaluate the result the way a structural biologist would: per-residue confidence, interface confidence, geometry validation and comparison against any experimental evidence available. Regions where the model should not be trusted are marked explicitly rather than smoothed over.
Key Features & Advantages
Complexes, Not Only Single Chains
Multimer and complex prediction covers protein–protein, protein–peptide, protein–nucleic acid and protein–ligand systems, which is where most project-relevant questions actually sit.
Confidence Read Correctly
pLDDT, PAE and interface scores are interpreted per region and per interface, distinguishing a genuinely disordered loop from a poorly modelled one — a distinction that changes what you do next.
Cross-Method Comparison
Independent predictors are run in parallel on demanding targets; convergence between architectures is stronger evidence than any single model, and divergence is reported.
Delivered Ready For Use
Models arrive energy-minimised, geometry-checked and prepared for docking or simulation, with alternative conformations supplied where the system is genuinely flexible.
Where This Service Is Used
1Obtaining a working structure for a target with no experimental coordinates
2Modelling a complex to understand how two partners engage before designing an inhibitor of the interaction
3Predicting the structural consequence of a clinical or engineered mutation
4Building antibody–antigen or receptor–ligand models to guide epitope-directed design
5Filling missing loops, termini and domains in an existing experimental structure
Methods and Techniques Applied
1MSA construction and depth assessment, which is often the single strongest determinant of prediction quality
2Monomer, multimer and complex prediction with current deep-learning systems, multiple seeds and recycles
3Per-residue and per-interface confidence analysis, including predicted aligned error interpretation
4Geometry validation: Ramachandran statistics, rotamer outliers, clashes and bond geometry
5Comparison with homologous experimental structures and with any client biophysical or crosslinking data
6Energy minimisation, protonation assignment and preparation of the model for downstream docking or MD
What We Need From You
1Sequences in FASTA for every chain in the assembly, with the intended stoichiometry
2Ligand, cofactor or nucleic acid identity where a complex is required (SMILES or sequence)
3Any experimental restraints available: crosslinking, mutagenesis, SAXS, low-resolution maps, known epitopes
4The downstream use — docking, MD, design or interpretation — since it determines how the model is prepared
What You Receive
1Predicted coordinates in PDB/mmCIF, with all ranked models rather than only the top one
2Confidence report: per-residue and interface scores, with regions unsuitable for design marked
3Validation summary covering geometry, clashes and agreement with experimental data where it exists
4Prepared structure for downstream work, including protonation and minimisation
5Annotated figures of the fold, the interfaces and the residues relevant to your question
What We Compute and Report
1Three-dimensional coordinates for every chain and for the assembly, with all ranked models
2Per-residue confidence values and inter-domain error estimates across the whole model
3Interface confidence for each contact surface in a complex, reported separately from global confidence
4Geometry validation statistics: Ramachandran distribution, rotamer outliers and clash score
5Structural deviation against homologous experimental structures where any exist
Questions We Are Asked Before Scoping
How accurate will the model be?
▸For well-covered single domains, typically close to experimental accuracy in the core. Accuracy falls for shallow alignments, disordered regions, unusual folds and flexible interfaces — the confidence report states which case applies to your target.
Can a predicted structure replace a crystal structure?
▸For many design decisions it is sufficient; for binding-mode certainty and for regulatory or publication claims it is not. We state which of your intended uses the model can and cannot support.
Will prediction capture the conformation my ligand stabilises?
▸Not reliably. Predictors tend toward the dominant conformation in the training data. Where an induced-fit state matters, we combine prediction with simulation or induced-fit docking.
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
The target, or the complex you care about, has no deposited experimental structure
A model is needed quickly to unblock docking, design or mutation interpretation
You need to know which regions of a model can carry a design decision and which cannot
Consider another service when
A close experimental template exists in the conformational state you need
Homology Modeling & Loop Refinement Binding-mode certainty is required for chemistry decisions or publication
Protein–Ligand Co-Crystal Structure Analysis 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.
A confident-looking model whose binding site sits in a low-confidence region
Site quality is assessed separately from global confidence, and we state plainly when a pocket is not modelled well enough to screen against.
The predicted conformation is not the one your ligand stabilises
Where induced fit matters, prediction is combined with simulation or induced-fit protocols and the limitation is stated up front.
Disordered regions modelled as ordered and then designed against
Low-confidence regions are marked explicitly in the delivered model and excluded from design recommendations.
A Worked Example
A receptor–peptide complex with no structural precedent
The situation
A programme needed the binding geometry of a peptide ligand at a receptor for which only the apo fold was known.
What the work involved
Complex prediction was run across multiple seeds with an orthogonal method for comparison, interface confidence was assessed per contact surface, and the top models were energy-minimised and simulated briefly to test stability.
What came back
Two distinct binding modes, one of which was stable in simulation and consistent with existing mutagenesis. Both were delivered, with the mutations that would discriminate between them.
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.
Around three quarters of predicted ligand poses fall within 2 Å
Current deep-learning structure prediction, evaluated on an independent protein–ligand benchmark, placed roughly 76% of ligands within 2 Å of the experimental pose, and reported approximately 80% by coordinate error — a substantial improvement over classical docking baselines on the same set, achieved without supplying a binding site. The corollary matters as much as the headline: roughly a quarter of cases still fall outside that threshold, which is why we report confidence per interface rather than delivering a single model.
Abramson et al., Nature, 2024 ↗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