Abstract:The overall purpose of this project is to enable logical inference of facts about health across species-boundaries and to identify the most effective species-specific interventions and drugs via the application of semantic web and machine learning.
Section I.
Creative Approach
Scientific Basis
Section II.
Experimental Plan
Time Plan
Essential Data
Outlook
What is your idea? Indicate in one or two sentences in bold the essence of your idea.
We are applying semantic web technologies and create ontologies to integrate and map knowledge on health as well as diseases, in particular age-related diseases, between animals and humans. Such a unifying description framework that bridges species boundaries will allow reasoning and inference of phenotypes and interventions and therefore will enable the transfer of knowledge between species boundaries.
Why is your idea an unconventional or creative approach to the problem outlined in the topic?
Aging and age-related diseases is a global issue which affects also poor and developing countries, but in these countries this research is very underdeveloped (as compared to Western countries). Furthermore, there is an increased susceptibility to diseases with age, for both communicable and incommunicable diseases, in both humans and animals, in both developed and developing countries.
Biomedical data on health and disease conditions are distributed in various formats, and knowledge on each species is often described with a specialized vocabulary and data structures. We are using advanced information technologies based on common web standards to unify knowledge from different species in order to make predictions about facts that are inferred from the combined knowledge on other species. This rational approach is based on the logic inference on graph structured data. The uniqueness of our approach is that we will apply machine learning methodologies for reasoning about the unified data and to make predictions that bridge the gaps between species concerning interventions against diseases.
Therefore the proposed research addresses the call for the combined metrics for measuring impact.
Describe the scientific basis for your idea and why you expect it to succeed.
An ontology is a formal structured representation of a number of concepts and the relations between them. They can be used to digitize knowledge and render this knowledge computer understandable. Ontologies are used in the semantic web to define the schema of data and are part of the knowledge representation of artificial intelligence. Ontologies contain inference and integrity rules, which enable the deduction of knowledge and guarantee of data quality.
Ontologies facilitate the integration of data and knowledge allowing interoperability and semantic information processing between diverse biomedical resources and domains as well as species. Thus, ontologies enable sharing knowledge between different organisms.
Comparing phenotypes between species potentially provides invaluable insights into human diseases. For instance, phenotypic characterization of mutants in different organisms can provide information that can be used to prioritize gene lists derived from gene expression profiling or genome-wide association studies. The ability to cross the species boundary has long been an extremely difficult problem, as human and model organisms' phenotypes are described using different formal ontologies and conceptual approaches. To change this, we are working to develop a series of ontologies and tools that use those ontologies, allowing for seamless integration of phenotypic data between species and predictions for effective treatments.
We will align phenotypes across species and enable the generation of a single, unified and logically consistent representation of phenotype data for multiple species.
The semantic web technology stack will be supplemented with fuzzy logic, network analysis and machine learning techniques.
This will enable the generation of logical predictions on healthspan extending interventions in animals as well as humans.
We are already actively working in this direction and have a team of scientists and developers.
How will you test it?
Describe your experimental plan, including any new technologies or tools to be developed.
We are using and developing advanced information technologies including semantic web and machine learning, and utilize concepts from systems and network biology to infer knowledge in other species.
First of all, we will develop semantic approaches to integrate ontologies. Our approach uses formal definitions of ontology classes to generate equivalences between phenotypes in species-specific ontologies. These formal definitions are generated using the phenotypes and train ontology together with multi-species or species-agnostic ontologies and gene ontologies, permitting the bridging between the ontologies. With the use of these cross-species ontology, we are integrating structured biological data into graphs. On these graphs we are applying machine learning algorithms to make predictions about healthspan extending interventions in various species and in particular in humans.
We will use fuzzy logic together with description logics, in order to determine the level of confidence for the suitability of a particular animal intervention for humans, and vice versa.
A core concept is the integrating of knowledge on aging and health from animals and projecting it to humans. For instance, to infer knowledge about human health from animals we also map homologous genes and construct molecular interaction networks. These interologs are used to complement human molecular interaction networks. In addition, coexpression networks and functional associations will be used to complement the networks. As a first proof of concept we are applying guilt-by-association methodologies to predict candidate genes associated with a certain disease process.
Gene candidates identified in humans will in turn be validated in suitable biomedical model organisms to assess their healthspan extending effects.
With the help of the ontologies we will also make predictions on small molecules and therefore identify drugs that have the potential to ameliorate diseases and extend the healthspan.
How will the work you describe be performed within the budget (USD$100,000) and time period (eighteen [18] months) allocated for the initial Phase I award? This 18 month time period should assume an approximate 12 month project work time with the additional six months to be used for ramp up and required reporting.
The first 6 months will be devoted to the integration of existing structured data and ontologies. Then, in the next 6 months, ontologies will be developed to complement the existing ones. With the help of those ontologies, data from various sources will be integrated. Then ontology will be populated with knowledge on individual entities like genes, proteins, processes and phenotypes. The constructed networks will be used to apply reasoning as well as network algorithms to infer promising candidates. The last 6 month will be used to test in silico predictions in lower model organisms as a proof of concept. Results will be published in peer-reviewed journals and presented at conferences.
What essential data will you generate during your Phase I award?
Overall, we will produce a unified resource description framework for health, which spans species boundaries. New ontologies for health data and disease will be created that are specific to cross species boundaries and utilize the knowledge in each species. New knowledge and facts will be inferred via logical reasoning, network algorithms and machine learning. Data on the effects of candidate genes and drugs on the healthspan will be generated to validate the approach. The source code will be released under open source license to let others contribute and reuse it for their research.
If your experiments in Phase I are successful, what are the next steps?
When our proof-of-concept experiments succeed we will utilize natural language processing to infer knowledge from the biomedical literature and enhance the predictive power of the constructed knowledge base. More advanced information technologies, especially those from the artificial intelligence field, will be used to make superior predictions.
The predictions will be tested in animals and those drugs that appear to be effective in multiple model organisms will be used for performing clinical trials in humans.
