Harnessing Data and AI to Advance Neuroscience Research

At Alkermes, patient-inspired science is the driving force. The company applies its deep expertise in neuroscience to develop medicines designed to help people living with complex and difficult-to-treat psychiatric and neurological disorders. By combining a robust portfolio of medicines, Alkermes is working to make a meaningful difference in how people manage their health.

That mission to push boundaries in drug discovery and development was the foundation of a recent Challenge Project with Break Through Tech’s AI Program. Together, Alkermes and a team of our AI Program Fellows explored how modeling and informatics can support critical stages of research, target identification, compound nomination, and the broader process of drug discovery.

The Challenge: From Millions of Data Points to Predictive Insights

Drug discovery often begins with massive datasets. In this project, the Alkermes team tasked students with using the ChEMBL database—containing over 2.4 million compounds, 1.6 million assays, and 15,000 targets—to generate predictive models and evaluate how confident those models are in their predictions.

This was no small task. The challenge required the Fellows to not only build accurate models but also design an interface that could assess prediction confidence, enabling scientists to better understand the reliability of the results.

Suggested Framework: Building Models with Confidence

To guide the Fellows, Alkermes outlined a suggested framework for approaching the project. The idea was to:

• Explore the use of compound structure-derived descriptors as independent variables for model building.
• Apply tools like Python and Scikit-learn, with the option to test deep learning frameworks.
• Consider conformal prediction methods to help quantify confidence levels in the models.
• Envision an interface capable of batch processing, making predictions and confidence assessments more accessible to researchers.

This roadmap gave students exposure to the types of methods and tools real-world scientists employ when working at the intersection of informatics and drug discovery.

The Impact: A Window into Real-World Drug Discovery

For the students, the Challenge Project was a hands-on opportunity to work on a problem that sits at the cutting edge of computational drug discovery. They gained practical experience in applying machine learning techniques to one of the most complex and high-stakes fields of science.

For Alkermes, the project reinforced the value of bringing diverse and emerging talent into the research process. By engaging the Fellows, the company explored innovative approaches to informatics while contributing to the education and growth of the next generation of AI professionals.

Looking Ahead

As Alkermes continues to pursue its mission of developing innovative medicines for psychiatric and neurological conditions—often in areas that have been overlooked or stigmatized, projects like this demonstrate the power of combining scientific expertise with fresh perspectives from rising technologists. By collaborating with Break Through Tech, Alkermes is not only advancing its own research but also helping to equip a new generation of AI talent with the skills, confidence, and real-world experience they need to shape the future of tech and healthcare.

This collaboration is a testament to what’s possible when industry leaders and emerging talent come together: a future where data, AI, and neuroscience converge to bring forward new solutions for patients who need them most.

Join us in creating Challenge Projects that deliver fresh, innovative solutions for your team, while giving Break Through Tech Fellows the real-world experience they need to launch careers in tech.