Grant: Discovery Grants Program - Individual
Université Laval
$36,000.00
Graph-based Natural Language Querying of Integrated Biological Knowledge Databases (KibioAI)
EF 2024-25
Résumé
Rationale. Currently, hundreds of biological knowledge databases (BKDs) exist to store biological data, and their integration is essential to get a holistic view of biological phenomenons. These databases are highly valuable to biologists and are essential to understand disease mechanisms at the molecular level. During a previous NSERC discovery grant, we developed approaches to handle and integrate these big biological databases and we built tools to better exploit them. The current way to use this data is using R tools, visualizations, and filters. However, biologists are not always familiar with many databases and their content, which currently limits the audience. A solution may lie in the advent of Large Language Models (LLMs) and conversational agents (i.e. chatbots), capable of analyzing very large amounts of text data, emulating natural language comprehension, and generating contextually relevant answers to questions, with ability to reveal hidden relationships based on extensive training. However, current conversational agents often lack trustworthiness because they produce incorrect information. This issue will be resolved by training LLMs using curated biological knowledge graphs, transformed into natural language.
Objectives and methodology. Our objective is to integrate many BKDs and train LLMs on this data. The initial step is to construct advanced biological knowledge graphs to improve BKD cohesion and extract valuable insights and knowledge using LLMs and conversational agents. To leverage graph structures with LLMs we will need to produce large amounts of relevant sentences describing relationships within BKDs, and train a LLM. The conversational agent, named KibioAI, will use the LLM model to enable users to interact with all BKDs at the same time using natural language queries, and will generate coherent and contextual answers with sources and explainability through graphs. The final step will involve evaluating KibioAI by benchmarking it against manually curated data and real-world use cases.
Impacts. Once rigorously evaluated, the developed conversational agent will facilitate meaningful dialogues to refine search and provide recommendations. This will enhance accessibility for biological researchers while addressing issues of trust and explainability in AI. The impact of this research aligns with AI's transformative potential in research. It will enhance the accessibility of integrated biological data, expanding its utility in research areas such as drug discovery, biomarker identification, and eventually personalized medicine. This initiative will also create a stimulating research and training environment, attracting talented HPQ, as it contributes to the digital transformation era. Finally, the proposed research will reinforce the university's research capabilities in AI-driven bioinformatics, fostering broader interdisciplinary initiatives and establishing Canada's presence in the evolving field of LLMs.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2024-06546
Name nom
Droit, Arnaud
Department département
Médecine Moléculaire
Organization id
51
Institution établissement
Université Laval
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
36000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
802
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Information systems and technology
Area of application fr
Systèmes d'information et technologie de l'information
Research subject code
2801
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Natural language and speech understanding
Research subject fr
Langage naturel et reconnaissance de la parole
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Graph-based Natural Language Querying of Integrated Biological Knowledge Databases (KibioAI)
Keywords
biological networks; graph analysis; natural language processing; conversational agents; generative ai
Application summary
Rationale. Currently, hundreds of biological knowledge databases (BKDs) exist to store biological data, and their integration is essential to get a holistic view of biological phenomenons. These databases are highly valuable to biologists and are essential to understand disease mechanisms at the molecular level. During a previous NSERC discovery grant, we developed approaches to handle and integrate these big biological databases and we built tools to better exploit them. The current way to use this data is using R tools, visualizations, and filters. However, biologists are not always familiar with many databases and their content, which currently limits the audience. A solution may lie in the advent of Large Language Models (LLMs) and conversational agents (i.e. chatbots), capable of analyzing very large amounts of text data, emulating natural language comprehension, and generating contextually relevant answers to questions, with ability to reveal hidden relationships based on extensive training. However, current conversational agents often lack trustworthiness because they produce incorrect information. This issue will be resolved by training LLMs using curated biological knowledge graphs, transformed into natural language.
Objectives and methodology. Our objective is to integrate many BKDs and train LLMs on this data. The initial step is to construct advanced biological knowledge graphs to improve BKD cohesion and extract valuable insights and knowledge using LLMs and conversational agents. To leverage graph structures with LLMs we will need to produce large amounts of relevant sentences describing relationships within BKDs, and train a LLM. The conversational agent, named KibioAI, will use the LLM model to enable users to interact with all BKDs at the same time using natural language queries, and will generate coherent and contextual answers with sources and explainability through graphs. The final step will involve evaluating KibioAI by benchmarking it against manually curated data and real-world use cases.
Impacts. Once rigorously evaluated, the developed conversational agent will facilitate meaningful dialogues to refine search and provide recommendations. This will enhance accessibility for biological researchers while addressing issues of trust and explainability in AI. The impact of this research aligns with AI's transformative potential in research. It will enhance the accessibility of integrated biological data, expanding its utility in research areas such as drug discovery, biomarker identification, and eventually personalized medicine. This initiative will also create a stimulating research and training environment, attracting talented HPQ, as it contributes to the digital transformation era. Finally, the proposed research will reinforce the university's research capabilities in AI-driven bioinformatics, fostering broader interdisciplinary initiatives and establishing Canada's presence in the evolving field of LLMs.