ABU DHABI 4th May 2023 (WAM). The Human Phenotype Project hackathon was held in Abu Dhabi by the Mohamed bin Zayed University of Artificial Intelligence and the Weizmann Institute of Science. The event brought 24 postgraduate researchers and students from different backgrounds together to tackle the most pressing healthcare challenges.
The term “phenotype” refers to the observable characteristics of an individual, including height, eye color, blood type and facial features. These traits are determined both by genetic makeup (genotype), and environmental factors.
The hackathon lasted two days and focused on analysing real data from a multi-omic deep-phenotype biobank in order to develop innovative solutions that address challenges of personalised and predictive analytics in healthcare. A deep-phenotype biobank is a collection biological samples such as saliva, blood or tissue that are linked with detailed clinical and health information about the individual from whom they were taken.
The hackathon will focus on WIS’s and Pheno.AI’s Human Phenotype Project in Israel, the world’s most comprehensive phenotype cohort. This project has seen the institute create a database with multi-omic and clinical measurements of health from more than 10,000 participants. The project’s deep profile includes each individual genome as well as other omics, meaning details about the constituents and functions of cells. It also includes information like medical history, lifestyle, nutritional habits, blood tests and continuous glucose and sleep monitoring.
The event was part MBZUAI and WIS’s ongoing partnership in order to promote AI and conduct both basic and application research on machine learning, computervision, natural language processing and computational biology. HPP goals include disease prevention and prediction, personalised treatments, and lowering drug development costs.
Participants included 14 postgraduates from MBZUAI and six from Weizmann Institute. Four other institutions were also represented. Pheno.AI trained the Hackathon participants on how to access the HPP data using the HPP trusted environment. It is the first instance that an international university was given access to these data. Participants worked in teams to apply AI-based techniques and develop different predictive models to predict values and phenotypes for a variety of data types.
Professor Eran Segal is the leader of HPP and a faculty at WIS and adjunct professor at MBZUAI. He said: “The HPP Hackathon places the participating students in the forefront of biomedical innovation, using cutting-edge analytical tools and data to make discoveries and develop solutions that can improve patient outcomes and care.” The event also served to remind us of the importance of collaboration by bringing together the best talents from MBZUAI, Weizmann Institute and other institutions in order to address global healthcare challenges.
Each of the six teams competing was given a choice between two challenges: a challenge on prediction and a challenge on creativity. The teams were required to predict chronological age based on biological markers for the predictive challenge. The winning team from MBZUAI came in first place by carefully choosing which phenotype data to use and which to ignore. They realized that the gut microbiome was too complex to be used for this task and focused instead on other data such as sleep monitoring and blood pressure measurements. The team also relied on machine learning models, such as gradient-boosted tree to deliver the best possible results.
A team from WIS was the winner of the creative challenge with the idea to extend the HPP database’s value to a broader audience by allowing people to gain insight into their own health. The idea of the team was to passively gather data from individuals. For example, photos taken by their phones showing what they eat. This data could be used to analyze their diet. The team used these insights to make predictions about these individuals’ health, including their risk of developing Type-2 Diabetes. They analysed them as though they were part the HPP dataset and had similar eating habits. This approach was shown to be a valuable way of predicting people’s Type-2 diabetes risk based on passive data.
Timothy Baldwin, MBZUAI’s Acting Provost said, “The UAE, and other parts around the world, could benefit from deep-phenotyped databases of human data that can help create a healthier and better future.” The HPP Hackathon brought together students, faculty and experts from different organisations. It gave them an opportunity to network with their peers and expand their skills, while tackling different health conditions and diseases.
It also showed the power of AI in analysing vast amounts of data to identify potential markers, which would otherwise not be possible at this scale. This work could lead to new discoveries about personalised and predictive analytics in healthcare, possibly paving the path for new treatments,” Baldwin added.
Mentors and experts provided guidance and support to the teams, including members of MBZUAI and WIS faculty, as well as Pheno.AI (the company that built the HPP platform, and managed the data collection for the project). The judging panel consisted of Professor Agathe Guilloux, Assistant Professor Martin Takac and Visiting Associate Prof. Tongliang Liu from MBZUAI’s Machine Learning Department, along with Dr. Hagai Rosman and Dr. Alon Diment Carmel who are senior researchers at Pheno.AI.