Watson for Drug Discovery is a cloud-based platform that aggregates diverse biomedical data sources to help researchers discover new disease pathways and drug targets. It uses natural language processing, predictive analytics, and dynamic visualizations. Case studies showed that Watson correctly predicted several known kinase interactions with the p53 protein and prioritized drug combinations for cancer immunotherapy. Researchers were able to generate novel hypotheses faster and more efficiently with Watson than traditional literature review methods.
Watson Corpus yield access to numerous types of public and private content to provide novel insights, dynamic visualizations, and ranked predictions.
Private Data
Public Data
Structured Data
Unstructured Data
Discuss/use cases for each type of data in the corpus
Cognitive platform that reads various types of content, learns through machine learning and expert training and evaluates through reasoning algorithms.
All built on the secure, HIPPA-compliant Watson Health Cloud.
How Watson for Drug Discovery Works
NLP:
Trained with domain-specific dictionaries, ontologies and subject matter experts
Understands semantic and contextual meanings
Understands the language of healthcare and life sciences
Predictive Analytics (formerly known as Reasoning Analysis)
Improves decision making reasoning algorithms and predictive models backed by evidence
Helps generate novel hypotheses by predicting potential relationships that may not be known
Visualization
Dynamic visualizations map detected connections between entities
Rich visuals allow for rapid learning
Interactive research experience utilizing various filters and views added or layered across different entities.
Cloud-based Platform *Future Iterations WDD will be supported by Cloud*
The secure, HIPAA-compliant platform was built with compliance needs in mind
Near real-time updates enable innovative interactions with data
The flexible health platform allows you to focus resources on solving new problems
Experts are able to see hidden connections between entities based on known features and properties discussed in the literature curated.
Sub graphs reveal clusters that correspond with common properties of interest
Filter views give researchers the ability to interact with visualizations by adding and layers different entities.
Filter views enable researchers to discover evidence-based answers to questions and to explore relationship on various levels and views.
Watson reads by using the following:
Dictionaries derive baseline meanings of entities and their synonyms
Ontologies help build relationships between entities
Entity Annotators categorize entities according to conceptual topics
Relational Annotators establish meaningful connections between entities and entity types
The Knowledge Graph maps every known relationship throughout the Watson corpus to deliver the network and of cause and effect relationships hidden within content
Specific rules – doesn’t just understand grammar; understands rules.
Watson reads by using the following:
Dictionaries derive baseline meanings of entities and their synonyms
Ontologies help build relationships between entities
Entity Annotators categorize entities according to conceptual topics
Relational Annotators establish meaningful connections between entities and entity types
The Knowledge Graph maps every known relationship throughout the Watson corpus to deliver the network and of cause and effect relationships hidden within content
Specific rules – doesn’t just understand grammar; understands rules.