Biodatathon: Empowering Drug Research with AI and Data Science

Prof. Dr. Swati Chandna, Head of Applied Data Science and Analytics (M.SC), was proud to organise the Biodatathon event in collaboration with Neuron Soultions and the Biomed X Institute. This exciting hackathon, which took place at the Data Lab on the Heidelberg Campus, brought together experienced participants to tackle challenging tasks in pharmaceutical research using the latest AI and data science techniques.

The challenge: revolutionising drug development
Drug development is a monumental task. On average, it takes 14 years to develop a new drug, with an estimated 90% failure rate. These figures emphasise the immense complexity, time and resources required for successful pharmaceutical innovation.

Objective of the Biodatathon

The Biodatathon was developed to explore how cutting-edge AI solutions can be utilised:

1. Achieve peak performance on question-and-answer benchmarks over biomedical knowledge graphs.
2. Improve understanding of how multimodal embeddings can improve the search for answers.
3. Demonstrate the potential of integrating these advances into an AI agent framework with tool access and a chatbot UI front-end to increase productivity when working with biomedical knowledge graphs.

Joint effort

The event was a true collaboration:

• SRH University, Heidelberg Campus: it hosted the hackathon and provided a lively academic environment. The main representative of the university was Prof Dr Swati Chandna, who heads the university's Applied Data Science and Analytics (ADSA) department and who actually came up with the idea for this hackathon.
• Biomed X Institute: Designed the challenging task that encouraged participants to push the boundaries of AI in drug discovery. Douglas McCloskey, who is leading the development of a Next Generation Virtual Patient Engine (VPE) for the clinical translation of drug candidates, and Gurdeep Singh, a bioinformatician, and other colleagues from their team were actively involved.
• Neuron Solutions: Provided mentors alongside the experts from the other organisers to guide the teams towards effective solutions and supported SRH University with general organisational tasks, with two of our co-founders, Lev Szabados and Otto Werschitz, present at the event.

Results: Science meets innovation

Three talented teams took part in the hackathon, each presenting innovative and realisable ideas. Work included multimodal embeddings for drugs and proteins, enrichment of nodes lacking features to improve insights, and subgraph extraction along with an agent-based workflow for enriching, embedding and re-ranking candidate answers. Their solutions showed real scientific value, and true to the collaborative spirit of the event, their work has been published as open source. These results illustrate the power of collective endeavour and the potential of AI to transform drug discovery processes.