Drug Binding Affinity Prediction Using Preprocessed Molecular Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current new drug development process is inefficient, with only a few pharmaceutical companies succeeding in developing new drugs due to the lengthy and costly process of candidate substance discovery and clinical trials, necessitating the use of digital technologies like artificial intelligence to enhance productivity.
Innovation Solution
A method and device that predict the binding affinity between proteins and compounds using preprocessing of compound and protein information, followed by reinforcement learning to determine the potential of compounds as new drugs, incorporating additional information such as ADMET properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional candidate substance discovery process is used, then comprehensive testing can be performed, but the development time exceeds 5 years and cost exceeds 2 trillion won
Solution Approach 1:
The patent applies preliminary action by pre-processing compound and protein information into feature vectors before actual binding affinity prediction. The system pre-processes molecular structures, extracts features, and creates embedded representations in advance, so that during the actual drug discovery process, the binding affinity can be quickly predicted without performing comprehensive从头 testing, thus reducing the 5-year discovery period while maintaining evaluation accuracy
Solution Approach 2:
The patent uses copying by creating computational models and simulations that replicate physical binding interactions. Instead of performing actual physical experiments for every candidate substance, the system creates digital copies of the binding process through machine learning models trained on existing data, allowing rapid virtual screening of thousands of compounds without time-consuming laboratory testing
2Reliability
If comprehensive preclinical and clinical trials are conducted, then drug safety and efficacy are ensured, but the total development time exceeds 10 years
Solution Approach 1:
The system performs preliminary action by conducting in-silico preclinical trials using the binding affinity prediction model before actual animal or human testing. The machine learning model predicts potential toxicity and efficacy issues in advance, allowing researchers to filter out problematic compounds virtually, thus reducing the need for extensive physical preclinical trials and shortening the overall development timeline while maintaining safety standards
Solution Approach 2:
The patent implements feedback mechanisms where the prediction model continuously learns from actual trial results. As real preclinical and clinical data becomes available, it feeds back into the system to refine and improve the binding affinity predictions, making the computational model progressively more accurate and reliable, thereby reducing the number of physical trials needed over time
3Device complexity
If only a few companies conduct new drug development, then resource concentration occurs, but productivity remains low with only 6% success rate
Solution Approach 1:
The patent applies universality by creating a binding affinity prediction system that can be applied across multiple drug discovery projects and by multiple organizations. The machine learning model serves multiple functions: predicting binding affinity, estimating toxicity, screening compounds, and prioritizing candidates. This universal tool can be deployed by any pharmaceutical company or research institution, democratizing access to advanced drug discovery capabilities and improving overall industry productivity
Solution Approach 2:
The system replaces mechanical (physical) experimentation with computational prediction. Instead of relying on labor-intensive wet lab experiments for every candidate evaluation, the patent substitutes these with automated computer-based binding affinity calculations using pre-processed molecular features and trained machine learning models, dramatically increasing throughput and efficiency
Data Source
AI summary
The present disclosure relates to a new drug predicting method, and device for performing method. A method for predicting new drugs includes generating preprocessed compound information by preprocessing compound information of a compound, by a new drug predicting device; generating preprocessed protein information by preprocessing protein information of a protein, by the new drug predicting device; concatenating the preprocessed compound information and the preprocessed protein information by the new drug predicting device; and predicting a binding affinity based on the concatenated preprocessed compound information and preprocessed protein information by the new drug predicting device.