Technology

Opyl’s TrialKey technology applies artificial intelligence to improve clinical trial design. It uses biostatistical validation, high-fidelity simulations, and predictive modelling to address inefficiencies in trial execution. The platform integrates real-world data with machine learning to improve trial success rates.

Data-Driven Trial Optimisation

TrialKey processes structured and unstructured clinical data from trial registries, protocols, and real-world sources. The platform extracts insights using natural language processing (NLP) and incorporates over 1,300 variables to refine trial design.

Key Capabilities

  • Pattern Recognition – Identifies trial success factors through clustering, dimensionality reduction, and association rule mining.
  • Causal Inference – Differentiates correlation from causation using propensity score matching and directed acyclic graphs (DAGs).
  • High-Fidelity Simulations – Runs up to 100,000 trial simulations to evaluate different scenarios, reducing risk in trial planning.

Explainable AI for Decision-Making

TrialKey uses explainable AI (XAI) to provide transparency in trial predictions. It applies Shapley values and feature importance rankings to quantify how variables impact trial outcomes. Expert validation ensures reliability and regulatory alignment.

Predictive Accuracy

  • 92%+ precision in identifying trials that will meet their primary endpoint.
  • Top 10% of predictions capture 42.5% of successful trials.
  • Top 20% of predictions capture over 80% of successful trials.

Data Collection and Processing

TrialKey gathers trial data from registries such as ClinicalTrials.gov and ANZCTR. The data is cleaned and structured using NLP models to generate meaningful trial features.

Variable Categories

  1. Trial Execution Variables – Recruitment numbers, inclusion/exclusion criteria, trial endpoints, principal investigators, and trial sites.
  2. Technology Variables – Drug mechanisms, administration methods, dosage, and treatment frequency.

Machine Learning Models

TrialKey applies generalised linear models, Bayesian inference, and ensemble methods to refine predictions. Probabilistic confidence intervals and sensitivity analyses help stakeholders assess uncertainty and variability.

Proven Results

TrialKey correctly predicted the success of the Moderna and Pfizer vaccines before clinical trials concluded. Its predictive power helps pharmaceutical companies, CROs, and investors make informed decisions.

Clinical Trial Variables

TrialKey analyses a wide range of trial variables to improve accuracy. These include:

  • Recruitment & Design – Patient numbers, treatment arms, primary and secondary endpoints, trial locations.
  • Regulatory Factors – Masking, blinding, intervention types, and allocation models.
  • Patient Criteria – Age groups, eligibility, health status, gender distribution.

Addressing Trial Challenges

High Failure Rates

Over 65,000 clinical trials run each year, yet 90% fail. Many failures are due to poor trial design, not the drugs themselves. TrialKey optimises trial setup by analysing large datasets rather than relying on limited biostatistical models.

Optimising Variables

TrialKey refines variables like patient recruitment, endpoint selection, and trial duration. Its AI-driven insights help researchers reduce costs and improve outcomes. A 15% improvement in trial success rates can save millions in drug development costs.

Advanced Data Modelling

TrialKey processes structured and unstructured data from 468,429 trials and 2.69 billion patients. It curates variables from both traditional datasets and free-text fields such as trial protocols and eligibility criteria.

Explainable AI and Reporting Bias

Poor trial reporting skews traditional prediction models. Less than 20% of trials fully report outcomes. TrialKey corrects for bias using data enrichment, linking across trial phases, and integrating press releases and publications.

Financial and Predictive Impact

TrialKey provides transparent probability distributions and confidence intervals for trial success. Investors, CROs, and pharmaceutical companies use these insights to reduce uncertainty in clinical development.

TrialKey’s Competitive Advantage

Curated Data & Proprietary Modelling

General AI models like ChatGPT cannot replicate TrialKey’s capabilities. TrialKey integrates:

  • A curated dataset from 468,429 trials, enriched beyond publicly available data.
  • Real-world evidence from 2.69 billion patients.
  • Explainable AI techniques tailored for clinical trial success prediction.
Enquire now