No Reagents for the In Silico Phase
Confidence-Scored Results
Publication-Ready Figures
Validation Planned With Partners
Response Within One Business Day
Service Overview
ICDMO's computational platform forecasts protein interactions and evaluates binding sites with confidence metrics, so that downstream experimental effort goes to the most plausible candidates. The service covers protein–antibody, protein–protein, enzyme–substrate and protein–nucleic acid interactions. Confidence scores narrow the candidate pool before you commit to expensive experiments; they do not establish that an interaction occurs. Experimental confirmation — for example yeast-based validation — is designed with you and coordinated through partner laboratories.
Technical Advantages
✓
No Experimental Material Required
The computational phase needs no reagents, cloning or assays from you — only the target sequence or structure. The analysis itself is of course a paid service.
✓
Faster Than the Bench, Scoped Per Project
Interaction maps and prioritised candidates arrive far sooner than an experimental campaign. The exact timeline is confirmed after scoping, since it depends on system size and input quality.
✓
Reliable Confidence Scoring
Confidence scores for each predicted interaction narrow the pool of candidates, directing experimental validation to the most promising pairs.
✓
Publication-Grade 3D Visualization
High-resolution 3D protein structure analysis with binding site details meeting top-tier journal publication standards.
✓
Comprehensive Interaction Coverage
Covers six interaction types: protein–antibody, protein–protein, protein–peptide, enzyme–substrate, 3-way protein interactions, and protein–nucleic acid.
✓
Optional Downstream Validation
Seamlessly paired with yeast two-hybrid, alanine scanning mutagenesis, and other experimental validation services.
Application Scenarios
1Protein–Antibody Affinity Prediction: Predict antibody binding to target epitopes before committing to immunization campaigns
2Protein–Protein Interaction (PPI) Modeling: Map interaction partners across the proteome, identify co-complex members
3Protein–Peptide Binding Affinity: Evaluate peptide inhibitors or substrate peptides against enzyme/receptor targets
4Enzyme–Substrate Binding Affinity: Predict substrate specificity, Km proxies, and binding pose for metabolic engineering
5Three-Protein Interaction Analysis: Model ternary complex formation critical for signal transduction and PROTAC design
6Protein–Nucleic Acid Interaction Analysis: Predict RNA-binding protein specificity, DNA-binding domain targets, TF–promoter binding
Downstream Experimental Validation Methodology
For predicted binding sites, ICDMO identifies candidate binding site amino acid sequences from protein coding nucleic acids. Each candidate site undergoes alanine mutation (alanine scanning), followed by yeast two-hybrid verification to confirm or refute the computational prediction. This orthogonal validation approach provides high-confidence experimental support for AI-predicted interaction interfaces — enabling confident reporting in publications and drug design programs.
What Published Work Reports
1On an independent protein–ligand benchmark, current deep-learning complex prediction placed roughly 76% of ligands within 2 Å of the experimental pose, and about 80% by coordinate error — materially better than classical docking on the same set, and achieved without being told where the site is (Abramson et al., Nature, 2024).
2The same benchmark shows that roughly one case in four still falls outside 2 Å. This is why we report confidence per interface and treat a predicted binding mode as a hypothesis for mutagenesis or structural confirmation rather than a result.
3Interface confidence, not global model confidence, is what predicts whether a modelled complex is trustworthy. We report the two separately for every complex delivered.
4Published results above come from independent groups and indicate what these methods achieve in benchmark conditions. They are not ICDMO project results and do not predict the outcome for any individual target.
What We Need From You
1Sequences in FASTA for every protein in the pair or complex, with species and isoform stated
2Structures (PDB/mmCIF) where they exist, including unpublished internal structures under confidentiality
3Nucleic acid sequences, peptide sequences or ligand SMILES where the interaction involves a non-protein partner
4Any experimental evidence that constrains the interface: mutagenesis, crosslinking, HDX, known epitopes
5The decision the prediction supports, since a screening triage and a publication figure need different levels of rigour
How the Analysis Is Performed
1Multiple sequence alignment construction and depth assessment, which strongly governs achievable accuracy
2Complex prediction with current deep-learning systems across several seeds, with interface confidence recorded
3Pocket and interface detection using independent geometric and energetic algorithms for cross-checking
4Per-residue energetic decomposition and computational alanine scanning to rank interface contributions
5Comparison against homologous complexes and against any experimental restraints you supply
6Optional short molecular dynamics to test whether the predicted interface is stable in explicit solvent
What You Receive
1Confidence-ranked interaction list with the score for every pair assessed
2Predicted complex structures in PDB format, with all ranked models rather than only the top one
3Binding site residue lists with per-residue contribution and hot-spot identification
4Publication-quality 3D figures and ready-to-open molecular viewer sessions for the prioritised interfaces
5A written assessment of which predictions are strong enough to act on and which are not
Interpretation and Limitations
A confidence score expresses how consistently the model builds an interface, not the probability that two proteins meet in a cell. Predictions are weakest for shallow or transient interfaces, for partners with little coevolutionary signal, and for interactions that depend on post-translational modification or on a third partner. Predicted interfaces are hypotheses for mutagenesis or biophysical confirmation, and deliverables are for research use only.
Questions We Are Asked Before Scoping
Can you screen many candidate partners at once?
▸Yes. Batch prediction across hundreds of pairs is routine and is the usual way to triage a candidate list down to the few worth testing experimentally.
What confidence score is high enough to act on?
▸There is no universal threshold. We report interface confidence alongside the structural evidence and state, for your specific system, which predictions we would be willing to design an experiment around.
Do you run the validation experiments?
▸No. ICDMO is a computational CDMO. We design the validation — which residues to mutate, which assay discriminates best — and the work is executed by your laboratory or a qualified partner.
Technical Workflow
1
Scientific Consultation
We discuss the objective, the decision the result has to support and the data you already hold. Nothing is quoted before that is clear.
2
Feasibility Review
We check that your inputs, target class and available evidence can support the approach, and say plainly where they cannot.
3
Technical Proposal
Scope, methods, deliverables, timeline and price are confirmed in writing, along with the limitations we expect to report.
4
Computation
Alignments are built, complexes are predicted across multiple seeds and interfaces are scored, with pocket detection and energetic decomposition applied to the ranked results.
5
Report and Follow-Up
You receive the ranked interactions, the structures, the interface analysis and an explicit statement of which predictions merit experimental follow-up.
Service Timeline & Deliverables
Is This the Right Service for You?
Choose this when
A candidate interaction list needs narrowing before experimental testing
The interface between two proteins must be characterised at residue level
A binding site needs identifying on a target with no ligand precedent
Consider another service when
A pocket for small molecules is the specific requirement
Binding Site & Druggability Assessment The partner is DNA or RNA rather than a protein
Protein / Nucleic Acid Interaction Model 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 confidence score read as a probability that two proteins interact in a cell
The score describes how consistently a model builds an interface, not cellular reality; validation experiments are named in every report.
Transient or modification-dependent interfaces modelled as if they were stable
Such systems are identified at scoping and either declined or reported with the limitation stated up front.
Hundreds of pairs screened and the top of the list treated as findings
Score distribution is examined, promiscuous entries are flagged, and selectivity within the matrix drives the recommendation.
A Worked Example
Narrowing 180 candidate partners to six worth testing
The situation
A pulldown experiment had produced a long candidate list with no way to prioritise it.
What the work involved
Complexes were predicted in batch with interface confidence recorded per pair, hot-spot residues were identified for the leading interfaces, and short simulations tested whether the modelled interfaces held.
What came back
Six pairs with high interface confidence and stable modelled contacts, delivered with the residues to mutate in a validation experiment that would discriminate real from predicted.
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 Process
1
Online Consultation
2
Solution Matching
3
Service Contract
4
AI Computation
5
Project Report
Note: All services are for research use only and not intended for diagnostic or therapeutic purposes.
Get a Custom Quote
Our scientific team responds within 24 hours with a detailed technical proposal and pricing tailored to your research goals.
Contact Us Online ConsultationStandard Deliverables
✓Detailed analysis report (PDF)
✓Raw data files & processed outputs
✓High-resolution publication figures
✓Interaction scoring tables (Excel)
✓Project summary presentation
Related AI Services
Molecular Dynamics SimulationDigital Library-to-Library ScreeningDigital Protein/Nucleic Acid Interaction ModelAI Multimodal Protein–Compound ScreeningAI Digital Library ConstructionVHH Antibody AI Screening & ValidationWhy ICDMO AI?
Computational CDMO Focus
Reproducible, Versioned Pipelines
Confidence & Limitations Reported
Response Within One Business Day