Drug Candidate Discovery Using Shared Chemical-Transcriptome Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing target-centered drug development process is time-consuming, costly, and has a low success rate, especially when target proteins are unclear or undruggable, making it difficult to develop new drugs effectively.
Innovation Solution
A machine learning model is used to locate embedding vectors for chemical structures and transcriptome changes in the same vector space, iteratively learning to minimize differences in transcriptome data before and after chemical compound administration, enabling the output of new drug candidates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a target-centered method is used to discover new drug candidates, then the discovery process follows a systematic approach with identified target proteins, but the development time increases to about 5.5 years or more and the success rate remains below 0.02%
Solution Approach 1:
The patent inverts the traditional target-centered approach by using phenotype-driven drug discovery. Instead of starting with a target protein and searching for compounds that bind to it, the method starts with observed phenotypic effects of compounds on cells or organisms, then identifies the underlying molecular mechanisms and target proteins. This reversal enables discovery of drug candidates without requiring prior knowledge of disease mechanisms or target proteins, significantly reducing development time while maintaining high success rates through empirical validation of phenotypic effects.
2Productivity
If high-efficiency screening is performed to find compounds that bind to target proteins, then more drug candidates are identified, but the verification process becomes extremely complex and time-consuming
Solution Approach 1:
The patent applies preliminary action by conducting high-throughput phenotypic screening first to identify compounds with desired biological effects, before undertaking complex mechanistic verification. The methodology pre-filters compounds based on observable phenotypic outcomes (cell viability, morphological changes, behavioral effects) using automated systems, then performs targeted verification only on promising candidates. This approach identifies hundreds of drug candidates efficiently while reducing verification complexity through staged filtering and prioritization based on phenotypic significance.
3Ease of manufacture
If the target protein hypothesis is essential for drug discovery, then the process can be systematically structured, but drug development becomes impossible when the target is undruggable or unknown
Solution Approach 1:
The patent implements universality by creating a phenotype-driven drug discovery platform that functions independently of target protein characteristics. The methodology uses universal phenotypic readouts (cell survival, proliferation, differentiation, behavioral changes) that can be applied to any disease state regardless of whether the underlying target is known, druggable, or amenable to traditional screening. This universal approach maintains systematic process structure through standardized phenotypic assays while achieving versatility across diverse disease mechanisms including undruggable targets like KRAS and complex diseases like Alzheimer's where target proteins remain unidentified.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure provides a new drug candidate material output apparatus, including: a communication module; a memory in which a new drug candidate material output program is stored; and a processor executing the new drug candidate material output program. The new drug candidate material output program provides a drug learning model in which an embedding vector for a chemical structure of a chemical compound and an embedding vector for change information on an amount of a transcriptome induced by each chemical compound are located in a same vector space, outputs a result of the change information on the amount of the transcriptome that matches the embedding vector for the chemical structure of the new material input to the drug learning model, or outputs information on one or more drugs that match the change information on the amount of the transcriptome that is a target input to the drug learning model.