Machine Learning MOA Clustering for Accurate PMR Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for predicting post-marketing requirements (PMRs) for pharmaceutical drugs are time-consuming, inaccurate, and subjective, making it difficult to determine whether a PMR will be required, which can lead to significant costs and inefficiencies in the drug development process.
Innovation Solution
A machine learning model is developed to predict PMRs by generating hierarchical tree structures of mechanism of action (MOA) data, applying frequent pattern mining algorithms, and clustering models to identify associations, using real-world data and regulatory data to provide probability scores for PMR predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional expert evaluation methods are used to determine PMR requirements, then the process can be performed with existing knowledge and guidelines, but the process is time-consuming and produces inaccurate, subjective results
Solution Approach 1:
The patent replaces the manual expert evaluation process with an automated machine learning system. The ML model processes clinical trial data, mechanism of action information, and regulatory guidelines to predict PMR requirements, eliminating the time-consuming and subjective nature of human expert evaluation while improving prediction accuracy and consistency
Solution Approach 2:
The system creates a digital replica of the expert evaluation process through training data that encapsulates regulatory guidelines and historical PMR decisions. The trained model then replicates expert judgment for new clinical trials, providing consistent and reproducible predictions without requiring actual expert involvement in each evaluation
2Productivity
If detailed regulatory guidelines are manually evaluated for each drug, then comprehensive coverage of regulatory requirements is achieved, but the process becomes inefficient and costly
Solution Approach 1:
The system performs preliminary processing by pre-training the machine learning model on historical data and regulatory guidelines before actual PMR predictions are needed. This preliminary action encapsulates complex regulatory logic in the trained model, allowing rapid prediction of PMR requirements for new drugs without manually re-evaluating all guidelines for each case
Solution Approach 2:
The patent extracts key regulatory requirements and historical PMR patterns from complex regulatory guidelines and historical data. These extracted insights are encoded into the machine learning model, separating the essential regulatory knowledge from the complexity of the original guidelines, thereby enabling efficient automated evaluation without losing regulatory comprehensiveness
Data Source
AI summary
A computer-implemented method for generating machine learning training data may include obtaining mechanism of action (MOA) data that is indicative of a hierarchical tree structure of relationships between the MOA data; generating linear representations of branches of the hierarchical tree structure; determining association rules for the MOA data by applying one or more frequent pattern mining algorithm to the linear representations; and determining, as at least a portion of the generated machine learning training data, MOA clusters by applying a clustering model to the linear representations and the association rules.