DILI Prediction Model Using Knowledge Graph Interaction Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing in silico techniques for predicting drug-induced liver injury (DILI) are limited by the scarcity of drug-induced gene expression data, leading to inefficiencies and unreliability in assessing the risk of DILI, as conventional methods require such data for adequate predictive performance.
Innovation Solution
A deep learning-based liver injury prediction model that utilizes molecular fingerprints, biological interaction profiles, and molecular properties of drugs to determine the probability of DILI, leveraging a knowledge graph to generate biological interaction profiles and combining them with molecular fingerprints and properties to enhance predictive performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional in silico techniques are used for predicting DILI, then predictive performance can be achieved, but the techniques are limited by the scarcity of drug-induced gene expression data
Solution Approach 1:
The patent introduces molecular fingerprints and biological interaction profiles as intermediary representations that mediate between the scarce gene expression data and the prediction task. These intermediaries capture essential drug characteristics and biological mechanisms without requiring extensive gene expression data, thereby resolving the contradiction between predictive performance and data scarcity
Solution Approach 2:
The patent transforms the prediction approach by changing from using raw gene expression data to using derived parameters such as molecular fingerprints (based on chemical structure) and biological interaction profiles (based on pathway analysis). This parameter transformation enables reliable predictions while accommodating limited data availability
2Reliability
If conventional methods are used for DILI prediction, then some predictive capability is maintained, but the methods are inefficient and unreliable due to data scarcity
Solution Approach 1:
The patent replaces the conventional mechanical approach of directly analyzing gene expression data with a computational approach using molecular fingerprints and biological interaction profiles. This substitution enables more efficient and reliable predictions by leveraging pre-computed molecular and biological representations that do not require extensive experimental data
3Reliability
If deep learning-based prediction model is used, then predictive performance is enhanced without relying on scarce data, but model complexity increases
Solution Approach 1:
The patent segments the prediction system into distinct functional components: molecular fingerprint generation module, biological interaction profile generation module, and prediction model module. This segmentation allows each component to process specific types of data independently, reducing overall model complexity while maintaining high predictive accuracy through specialized processing of molecular and biological information
Data Source
AI summary
A method may include determining, based at least on a knowledge graph, a plurality of biological interaction profiles associated with a plurality of drugs. The knowledge graph being representative of a plurality of interactions between a variety of drugs, proteins, and a hierarchy of biological functions. Each biological interaction profile may be representative of the effects of a corresponding drug being propagated through protein-protein interactions and biological functions. A liver injury prediction model may be trained, based on a training dataset including the biological interaction profiles, a probability of drug induced liver injury. The liver injury prediction model to may be applied to determine, based on the biological interaction profile of a drug, the probability of liver injury associated with the drug. In some cases, the liver injury prediction model may further determine the probability of liver injury based on the molecular fingerprint and/or the molecular properties of the drug.