Lifelike - an AI based knowledge mining platform accelerating research

Lifelike is a graph-powered knowledge consolidation and mining platform tailored to bio-sciences. Researchers can understand structured and unstructured data quickly, cross-connect it with other sources of knowledge, and with the help of graph-powered machine learning, provide interpretable and evidence based insights, to solve complex scientific problems and develop new treatments.  

Evelyn Travnik, Chief Information Officer, Center for Biosustainability

A global health emergency challenges the scientific and clinical community to swiftly characterize the threat, identify therapeutics and develop vaccines. Adding to the difficulty is the complexity of the domain, the influx of new and possibly contradictory information and the sheer number of people involved. 

The response to the current pandemic resulted in a huge explosion of new  information, amplifying an already existing problem, making it impossible to consume and comprehend in a reasonable timeframe. 

Knowledge is often expressed in content types such as PDFs, presentation slides and images, or spreadsheets and highly complex data schemas, making it difficult for humans and machines to process and enable data-driven learning.

Data and knowledge exists also at different confidence levels and we don't have sufficient means to filter it. For example, experiments done in a mouse might not translate to human, or clinical data for one ethnic group might not translate to another. 

In order to respond to a new threat effectively, many different disciplines  have to collaborate, new teams are formed, yet a communication gap exists between researchers, clinicians and data scientists, that when overcome can accelerate a new threat response significantly. 

Lifelike enables researchers, clinicians, data scientists and machines to work together efficiently to capture and process complex knowledge from a variety of disciplines. Groups of researchers may come from fields as diverse as genomics, drug discovery, clinical research and data science, yet with Lifelike they can integrate the knowledge in a format that they can all understand and communicate across.

More importantly, Lifelike's knowledge representation format can leverage graph-powered machine learning to process large volumes of interconnected data, enabling scientist and clinicians to move from big picture view to detailed mechanistic explanations, at the same time filtering knowledge (e.g. for a particular ethnic group) for targeted mining. This enables the development of new solutions that are explainable and evidence-based.

Stakeholders have the need to establish a 360-degree view in a short period of time. Knowledge is however spread over many data sources and content types, each providing partial answers to questions, representing only “one piece of the puzzle”. Lifelike provides an integrated knowledge base, enabling users to collaboratively piece together the full picture while relying on algorithms to point them in the right direction and detangle complexities.

Lifelike is an open source platform that will be available to the scientific community. As such, it will be able to support the teams and activities that we rely on to respond to global health crises.  

A fully integrated graph-powered knowledge mining platform, such as Lifelike, does not exist today as open source. The capabilities range from natural language processing, knowledge reconstruction (automated and manual), graph visualization as well as full-text and graph-powered search.

The platform contains pre-loaded public data sources combined into one integrated knowledge base that is updated regularly. Users can upload their own content (text or structed data tables) and contextualize it against the integrated knowledgebase to gain a full-picture view across many data sources. 

Knowledge reconstruction capabilities enable researchers across different disciplines to create and share new knowledge at various levels of sophistication, from taking visual notes, to creating complex multi-media data stories, to sophisticated knowledge models that can be analyzed with graph algorithms.

While the users of our system would be scientists, researchers, clinicians, and data scientists, the impact extends to the patients. 

Researchers and clinicians will be able to learn faster from emerging data by being able to contextualize it against prior knowledge from many sources. Since knowledge can be filtered based on evidence and other meta data, (e.g. genes/diseases prevalent in an ethnic, gender or age group), Lifelike can be used to develop targeted treatments for patients and improve outcomes.   

We are collaborating with a research center within the Greater Copenhagen Health Science Partners to validate our application for research and clinical applications. 

We plan to make our application available to as many researchers and scientists as possible and will encourage and facilitate knowledge sharing and crowdsourced knowledge curation.  

Lifelike is a hosted cloud solution that can be scaled as the audience grows. We plan to establish a consortium and partner with more research centers as well as commercial companies to further develop the platform. We have also been contacted by a large pharma company who is currently evaluating our solution for the use by 2000 researchers at a global level. 

The level of success will be measured by the number of active users, level of knowledge creation within groups and shared with the public and the number of institutes and companies we partner with. 

We have also performed several internal validation studies and collected success stories to demonstrate that our technology works which will drive further adoption. 

Novo Nordisk Foundation Center for Biosustainability, with locations as Technical University of Denmark and the University of California San Diego. 

We are looking to expand Lifelike to external users and are looking for exposure and partnership opportunities.