AI Model Predicting Human Substance Actions via Animal Biomarker Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting the effects and adverse events of candidate substances in humans during drug development are inefficient due to reliance on speculative biological reaction mechanisms, leading to high dropout rates and significant financial losses, as they require extensive information and are prone to incorrect predictions if the initial mechanism is flawed.
Innovation Solution
An artificial intelligence model is trained using data on the dynamics of biomarkers in non-human animals to predict the actions of test substances in humans, leveraging existing substances' known actions and biomarker data to improve prediction accuracy without relying on therapeutic mechanism speculation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods use speculative biological reaction mechanisms to predict substance actions, then prediction can be performed with available information, but prediction accuracy deteriorates due to potential mechanism errors and high information requirements
Solution Approach 1:
The patent creates a virtual copy of human biological response by training an AI model on animal biomarker data and human clinical outcomes. The model learns to map animal biomarker dynamics to human substance actions, effectively copying human response patterns from animal data without needing to understand or speculate about the underlying biological mechanisms.
Solution Approach 2:
The patent replaces the mechanical system of biological mechanism speculation with an data-driven AI prediction system. Instead of using logical construction of biological pathways, the system uses machine learning to directly predict human substance actions from animal biomarker data, substituting mechanism-based reasoning with pattern recognition.
2Reliability
If extensive information is collected for mechanism-based prediction, then comprehensive analysis is possible, but prediction efficiency deteriorates due to complex processing requirements
Solution Approach 1:
The patent changes the parameters of the prediction system by shifting from mechanism-based parameters (biological pathways, molecular interactions) to data-driven parameters (biomarker dynamics, statistical patterns). This parameter transformation enables the system to achieve comprehensive prediction accuracy without the computational burden of processing extensive mechanistic information.
3Reliability
If preclinical trials are conducted to evaluate candidate substances, then safety and efficacy can be assessed, but development time and cost increase significantly
Solution Approach 1:
The patent performs preliminary action by using the AI prediction model to evaluate candidate substances before conducting full preclinical trials. The model provides early indicators of potential efficacy and safety issues by analyzing animal biomarker data, allowing researchers to prioritize candidates more effectively and reduce the number of substances requiring extensive preclinical testing.
Solution Approach 2:
The patent introduces an intermediary AI prediction system between animal testing and human clinical trials. This intermediary uses biomarker dynamics to translate animal data into predictions about human substance actions, providing a bridge that reduces reliance on lengthy preclinical trials while maintaining evaluation accuracy.
4Measurement precision
If multiple existing substances are tested to build prediction models, then model accuracy improves through more training data, but data processing complexity increases
Solution Approach 1:
The patent creates a universal AI prediction model that can handle multiple substances and multiple biomarkers simultaneously. The model is designed to process diverse data types (different substances, different organs, different biomarkers) through a unified framework, reducing the need for separate processing pipelines for each substance or biomarker combination.
Data Source
AI summary
Actions, such as effects and adverse-events, of a test substance in humans are predicted by using an artificial intelligence model trained by a method for training an artificial intelligence model, the method including inputting into the artificial intelligence model a set of first training data and second training data or a set of the second training data to train the artificial intelligence model.