Biomedical Entity Identification Using Context-Rich Text Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current knowledge graph-based methods for identifying biological targets in drug discovery are limited by the loss of contextual information during data translation and the inability to query complex biological contexts, leading to inefficiencies and high failure rates in target identification.
Innovation Solution
A machine learning model is trained on entity-linked text sequences to predict unique biological entity identifiers directly from free text, allowing for nuanced queries and leveraging rich contextual information, overcoming the limitations of knowledge graphs by using language models to identify multiple relevant targets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If knowledge graph-based methods are used to identify biological targets, then data can be structured and processed systematically, but contextual information is lost during data translation
Solution Approach 1:
The patent introduces an intermediary representation layer between the raw biomedical text data and the final target identification output. This intermediary layer preserves contextual information by maintaining the semantic relationships and nuances of the original text while still enabling systematic processing through structured representations, thus resolving the contradiction between systematic processing and information preservation.
Solution Approach 2:
The patent employs a nested structure where multiple levels of information representation are embedded within each other. The contextual information from free text is nested within structured data representations, which are in turn nested within the target identification framework. This allows contextual information to be preserved at inner levels while enabling systematic processing at outer levels.
2Ease of manufacture
If knowledge graph-based methods are used, then data can be organized in networks with relationships, but complex biological contexts cannot be queried effectively
Solution Approach 1:
The patent creates a multi-functional system that can handle both structured data organization and complex natural language queries. The model is designed to perform multiple functions: organizing data systematically like a knowledge graph, while also understanding and querying complex biological contexts using natural language, thus providing universal adaptability across different query types.
Solution Approach 2:
The patent introduces dynamic query processing capabilities that allow the system to adapt to different types of biological context queries. Rather than being restricted to fixed knowledge graph schemas, the system dynamically interprets and processes varied query formats, enabling flexible adaptation to complex biological contexts while maintaining systematic data organization.
3Measurement precision
If manual analysis is used to assess biological target characteristics, then multiple properties can be reviewed, but the process is time consuming and costly
Solution Approach 1:
The patent replaces the mechanical manual analysis process with an automated machine learning system. The model automatically assesses multiple properties of biological targets by processing training data and applying learned patterns, thereby maintaining the precision of comprehensive property assessment while eliminating the time-consuming and costly manual review process.
Solution Approach 2:
The patent performs preliminary action by pre-training the model on extensive biomedical data before actual target identification. This preliminary training phase allows the model to learn and encode multiple property assessment criteria in advance, so that during actual use, the model can rapidly evaluate new targets without requiring time-consuming manual analysis of each property.
4Reliability
If more biological targets are screened to find high quality candidates, then the failure rate may be reduced, but the complexity of analysis increases
Solution Approach 1:
The patent extracts and prioritizes the most critical properties and features that determine target quality from the comprehensive set of possible characteristics. By identifying and focusing on the key discriminative features through training on labeled data, the model can effectively screen a large number of targets without requiring complex analysis of all possible properties, thus maintaining high success rates while managing analysis complexity.
Data Source
AI summary
A computer-implemented method of training a machine learning model to identify biological entities for drug discovery is disclosed. The method comprises providing a training data set comprising a plurality of entity-linked text sequences, each text sequence including a mention of a biological entity, where the biological entity is linked to a corresponding biological entity identifier from a set of possible biological entity identifiers; masking the mention of the biological entity within each text sequence; encoding each masked text sequence into an input representation for a machine learning model; and training a machine learning model to predict the unique entity identifier of the masked biological entity based on the input representation. The described method is able to utilise the full breadth of the rich contextual information available in the biomedical text corpus to predict new biological targets for drug discovery and avoids the restrictions intrinsic to relationship prediction using knowledge graphs. The ability to identify more promising, biologically relevant targets in an automated manner, significantly reduces the requirement of human input and reduces the failure rate in targets that are progressed in the drug delivery pipeline.