David Hodgson and Szymon Jakobsze, congratulations on your successful Ideathons. Can you tell us a little bit about why you decided to participate?

David:  I had never heard of an Ideathon, but my manager, Adam Kucharski, brought it to my attention. I didn’t know what to expect. However, some of the potential challenges you could choose heavily overlapped with my and some of my colleagues’ research, namely modelling correlates of protection for infectious diseases, so I was interested to give it a go and see how other groups would approach the problem. 

Szymon: For me it was an excellent opportunity to test my skills in data science and learn something new. For most of us, the topics were new and exciting, especially Natural Language Processing (NLP), and it was fascinating to learn something new and compare ourselves with other MSc and PhD students in the field.

What challenges did you have to complete for the Ideathon?

David: Two weeks before the Ideathon, we were given a large set of data and an outline of the challenge. The data included measures of various immune biomarkers (antibody titres, cytokine concentrations, genetic information) pre- and post-vaccination for Influenza and Pneumococcus for several individuals. We were asked to use this data to develop a platform for discovering and analysing correlates of protection against infectious disease. We then had three full days at the Wellcome Trust to help refine the platform whilst interacting with the other teams. It was cool to spend this time in the Wellcome Trust and see how everything works from the inside.

Szymon: Our challenge was to develop an NLP pipeline for vaccine uptake sentiment analysis based on social media data. NLP in machine learning is used to gain insights into the human language, for example determine the sentiment polarity of free text. We were asked to develop an NLP pipeline which would allow us to analyse sentiment towards vaccine uptake on Twitter to better understand the association with real-time vaccine uptake. Another point of the challenge was to include predictive methods to flag when a critical mass of ‘negativity’ has been reached to predict short-term and long-term ‘shocks to immunisation systems’ due to a decrease in real-time vaccine uptake.

Who did you work with?

David: Our team was made up of colleagues from LSHTM, Tim Russell and Elizabeth Williamson, as well as Gavin Kelly from the Francis Crick Institute.

Szymon: Our team was assembled from Health Data Science course participants who found the challenge intriguing and were willing to allocate time from their MSc projects. Ultimately, I teamed up with some great people: Gabriel Battcock, Oliver Dolin, Walter Muruet Gutierrez, and Dzan Ahmed Jesenkovic.

What did you come up with to approach your challenges?

David: We spent a long time cleaning the data as it was very messy! Once we got our heads around that, we applied some of our existing mathematical and statistical techniques to the cleaned data. We managed to simulate the post-vaccination antibody trajectories of some individuals whilst also applying machine learning algorithms to group individual cytokine responses into several categories. We then presented these results in a closed presentation to the Wellcome Trust judges. We demonstrated the potential impact that a combined platform (i.e one that can clean, analyse and plot immune biomarker data) could have on helping identify correlates of protection against infection and disease, particularly for infectious diseases where Phase 3 clinical trials are difficult to conduct. We emphasised the importance of such a platform in resource-limited settings that may need more academic resources to run the modelling and statistics in-house. 

Szymon: For our presentation, we created an app called "HORTON," named after the elephant from "Horton Hears a Who!" This app aimed to analyse Twitter data to identify tweets expressing doubts about vaccines, including safety concerns or mistrust in health organisations. Our goal was to monitor these sentiments in real-time and observe how they varied across different regions and demographics. We suggested that researchers could use this tool to gain a better understanding of the attitudes driving vaccine hesitancy and engage with these communities to develop public health interventions for increased vaccine uptake.

You were up against quite a few teams, why do you think you were so successful and what do you hope to do with your winnings?

David: Wellcome offered us a £100,000 starting grant to build the platform over the next 12 months. We are looking to hire a Research Software Engineer in the coming weeks to help us build a robust and user-friendly platform that other groups can use. Winning the prize is great, I think the domain-specific knowledge that each of us brought to the project made it possible to contextualise the work with the global research sphere better; I think Wellcome looks at this very favourably and we worked well as a team, which was picked up by the judges and helped us secure the win.

Szymon: I’d say what distinguished us from the other groups was the fact that we tried not to focus too much on making the most technically sophisticated solution, instead we decided to design the tool in the way that would be beneficial for the researchers but also remain non-invasive and engaging for the public users. Participating in the Ideathon served as inspiration for our independent research during our MSc studies. It helped us stay motivated and engaged with our academic work. The student prize, while not a research grant, has further encouraged us to pursue our academic interests. We will maintain all of our work on GitHub, and we’re open to potential future developments.