Workpackage 06

Workpackage Objectives

• Develop a new multivariate machine-learning based classification approach that is tailored for the use of imaging for drug development.
• Provide a user-friendly interface (Matlab Toolbox) to facilitate the use of the new analysis methods in industry/academia and collaborative projects.
• Extend the use of the classification methods to multimodal (structural/function MRI; MRI/genetic; MRI/neuropsychological) data. Extend the use of the classification methods to multimodal (structural/function MRI; MRI/genetic; MRI/neuropsychological) data.

Workpackage Leads

Academic Lead: Prof. Michael Brammer, King’s College London, UK
EFPIA Lead: Dr. Lori Badura, Pfizer Limited, USA /UK

Workpackage Partners

• H. Lundbeck A/S
• AstraZeneca AB
• Eli Lilly and Company Ltd
• GlaxoSmithKline Research and Development Ltd.

Synergy between academia and industry

Historically, almost all functional and structural MR tools that are currently widely used have been developed in academia to answer questions posed by academia (mainly concerning localisation of changes in brain responses or structure). To the best of our knowledge no project has ever involved joint industry development of tools optimised for use in drug discovery. While there are isolated papers about using machine learning methods to analyze neuroimaging data, there is currently no wide-scale application of machine learning focused on addressing the unique challenges of early drug development (though Pfizer, for example, do have in-house expertise in machine learning methods). NEWMEDS provides a focused academic-industry collaboration to expedite methods development and delivery of usable tools. Prof. Brammer’s team will contribute their skills in developing machine learning methods for MRI analysis and MRI/genetic analysis. Pfizer, AstraZeneca and Eli Lilly will contribute datasets specifically chosen to challenge the methods and foster their development – and would subsequently test the tools on their in-house early development projects and provide feedback regarding tool utility. In addition, it provides a synergy among industry partners in terms of driving consensus on the interpretation of neuroimaging data. The collaboration will progress rapidly because (a) it will have clear value in drug-development and testing using neuroimaging, and in the longer term, other data, (b) it will provide a critical, delivery-orientated testing environment, (c) there will be a methodological interchange with drug-industry machine learning expertise that will aid algorithm development, and (d) the academic arm will have access to ideal datasets for testing and development.