Federating and integrating what we know about the brain at all scales: Computer science meets the clinical neurosciences

Abstract

Our everyday professional and personal lives are irrevocably affected by technologies that search and understand the meaning of data, that store and preserve important information, and that automate complex computations through algorithmic abstraction. People increasingly rely on products from computer companies such as Google, Apple, Microsoft and IBM, not to mention their spinoffs, apps, WiFi, iCloud, HTML, smartphones and the like. Countless daily tasks and habits, from shopping to reading, entertainment, learning and the visual arts, have been profoundly altered by this technological revolution. Science has also benefited from this rapid progress in the field of information and computer science and associated technologies (ICT). For example, the tentative confirmation of the existence of the Higgs boson (CMS Collaboration et al. Phys Lett B 716:30-61, 2012), made through a combination of heavy industrial development, internet-based scientific communication and collaboration, with data federation, integration, mining and analysis (Rajasekar et al. iRODS primer: integrated rule-oriented data system. Synthesis lectures on information concepts, retrieval, and services. Morgan & Claypool, San Rafael, 2010; Chiang et al. BMC Bioinformatics 12:361, 2011; Marks. New Sci 196:28-29, 2007), has taken our understanding of the structure of inorganic matter to a new level (Hay et al. The fourth paradigm: data-intensive scientific discovery. Microsoft, Redmond, WA, 2009). But within this vision of universal progress, there is one anomaly: the relatively poor exploitation and application of new ICT techniques in the context of the clinical neurosciences. A pertinent example is the genetic study of brain diseases and associated bioinformatics methods. Despite a decade of work on clinically well-defined cohorts, disappointment remains among some that genome-wide association studies (GWAS) have not solved many questions of disease causation, especially in psychiatry (Goldstein. N Engl J Med 360:1696-1698, 2009). One question is whether we have the appropriate disease categories. Another factor is that gene expression is affected by environmental and endogenous factors, as is protein function in different circumstances (think of the effects of age, developmental stage and nutrition). It is clear that any genetic associations with disease expression are likely to be highly complex. Why then are the world’s most powerful supercomputers not being deployed with novel algorithms grounded in complexity mathematics to identify biologically homogeneous disease types, or to understand the many interactions that lead to the integrated functions that arise from DNA metabolism, such as cognition? Is it from a lack of appropriate data and methods or are the reasons related to our current clinical scientific culture?.