HomeServicesBioinformatics & Multi-Omics AnalysisClinical Prediction Modeling & Public Database Analysis
Ref H14
Computational Service
Confidence Reported
Scoped After Feasibility Review
Service Overview
Clinical prediction models are easy to build and hard to build well. We develop statistical and machine-learning models on clinical cohorts or public datasets with the methodology that makes them credible: pre-specified outcomes, principled handling of missing data, sample-size assessment, internal validation by bootstrapping or nested cross-validation, calibration assessment and decision curve analysis. Reporting follows established prediction-model guidance, so the work stands up to methodological review.
Key Features & Advantages
Calibration Reported, Not Only Discrimination
A model with good area under the curve can still produce badly wrong probabilities. Calibration plots and metrics are delivered alongside discrimination as standard.
Missing Data Handled Principledly
Multiple imputation with the missingness mechanism assessed is used rather than complete-case analysis, which routinely biases clinical models.
Optimism Corrected
Internal validation uses bootstrapping or nested cross-validation to correct optimism, so the reported performance is not the performance on the data the model learned from.
Clinical Usefulness Assessed
Decision curve analysis shows whether the model would actually improve decisions at plausible thresholds, which statistical performance alone does not establish.
Where This Service Is Used
1Developing a diagnostic or prognostic risk model from a clinical cohort
2Validating an existing published model in a new population
3Analysing public clinical and genomic databases for survival or outcome associations
4Building machine-learning models on structured clinical data with interpretable outputs
5Producing the statistical analysis and figures for a clinical research manuscript
Methods and Techniques Applied
1Data cleaning, variable definition, outlier handling and missing data assessment with multiple imputation
2Sample size and events-per-variable assessment relative to model complexity
3Model development with regression and machine-learning approaches, with penalisation where appropriate
4Variable selection with stability assessment rather than single-pass stepwise procedures
5Internal validation by bootstrapping or nested cross-validation, with external validation where data permits
6Calibration, discrimination, decision curve analysis and interpretability analysis such as SHAP
What We Need From You
1The clinical dataset with a data dictionary and the outcome definition
2Candidate predictors and any clinical constraints on which variables may be used in practice
3The intended use of the model: screening, prognosis, treatment selection or research only
4For public database work, the database and cohort definition, or a request that we propose one
What You Receive
1Cleaned analysis dataset with a documented derivation record
2Final model with coefficients or the serialised model object and the prediction equation
3Performance report covering discrimination, calibration, decision curves and validation results
4Interpretability analysis showing which variables drive predictions and in which direction
5Manuscript-ready tables, figures, methods text and reproducible analysis code
What We Compute and Report
1Model coefficients or the prediction equation, with the derivation fully documented
2Discrimination and calibration metrics corrected for optimism through internal validation
3Decision curve analysis across clinically plausible thresholds
4Variable importance and per-prediction explanation output
5Missing data assessment with the imputation approach and its sensitivity
Questions We Are Asked Before Scoping
Will machine learning outperform logistic regression?
▸On structured clinical data with moderate sample sizes, usually not by much. We fit both and report the comparison rather than assuming the more complex method wins.
Can the model be used clinically?
▸Not from this work alone. Clinical use requires external validation, impact assessment and, in most jurisdictions, regulatory clearance. Deliverables are for research use only.
What if our sample size is too small?
▸We say so at scoping and calculate what the data can support. Building a complex model on insufficient events produces a confident-looking result that fails on new data, which helps no one.
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 prognostic or diagnostic model must meet methodological reporting standards
An existing published model needs validating in a new population
Public clinical or genomic databases need analysing reproducibly
Consider another service when
Imaging features are the predictors
Medical Imaging, Radiomics & Pathomics Published trial evidence must be synthesised instead
Meta-Analysis & Mendelian Randomization 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.
Complete-case analysis quietly biasing the model
Missingness is assessed and multiple imputation applied, with sensitivity analysis on the assumption.
Model complexity exceeding what the event count supports
Events per variable are computed before modelling, and we report when the data cannot support the intended model.
Discrimination reported without calibration
Calibration plots and decision curve analysis are delivered as standard, since a well-discriminating model can still give wrong probabilities.
A Worked Example
A machine learning model that regression matched
The situation
A team wanted a gradient-boosted model for outcome prediction on 640 patients with 92 events.
What the work involved
Both a penalised regression and the ensemble model were fitted under identical nested cross-validation, with calibration and decision curve analysis computed for each.
What came back
Discrimination was equivalent within confidence intervals, while the regression calibrated better and was interpretable. The simpler model was recommended, with the comparison documented for reviewers.
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.
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