A data-driven approach to mental illnessNovember 15, 2019
Clinical translation of genetic loci information from genome-wide association studies into viable models needs a deeper understanding of the underlying molecular and biological processes. This, in turn, requires an integrative effort from clinicians, mental health researchers and those skilled in data science. One of the most interesting aspects of data science is the opportunity to adapt to emerging technologies for the generation of new links between phenotypes, neuro-images and other biological datasets. As mental illnesses like depression are often undiagnosed, an algorithm developed with artificial intelligence using clinical records and other data can help in early detection and intervention, potentially saving lives. A similar, data-driven approach has been very useful in developing a risk score called QRISK2 for cardiovascular events, incorporating data from 2.3 million data records in the United Kingdom. This risk algorithm assessed several parameters like blood pressure, smoking status and cholesterol, and in the process, demonstrated great collaboration between clinical, research and data science communities. QRISK2 is now commonly used by general practitioners, enabling them to discuss with patients strategies to reduce their risk of CVDs. In order to develop such algorithms, there is a need to develop a proper biobank with a given dataset. In India, there are several independent biobanks for various diseases. National Institute of Mental Health and Neurosciences, Bengaluru has taken a huge step in building a biobank for mental disorders. However, being a country with a large and diverse population, there is a need to develop much larger cohorts with integrated electronic health records. Thus, it is crucial for more clinicians to engage in such activity. Furthermore, in order to flag particularly vulnerable people in high-risk populations, data from outside the clinical context, such as that from social media, may play a potential role in improving algorithms and mental health outcomes. However, using people’s data without explicit consent needs careful consideration as users of social media may not be comfortable for even their public posts to be used in such a manner. In the recent launch of a project called “BHARAT”, the National Brain Research Centre has started work on a big data analytics-based Alzheimer’s Disease (AD integrating non-invasive magnetic resonance imaging (MRI), MR spectroscopy and neurophysiological test outcomes. The big data collection in this project is aimed at facilitating multi-modal data analysis for feature extraction, classification and decision fusion.
One more interesting aspect of data science is that linking routinely collected health data to the data set could provide new clues for repurposing existing therapies for new indications. There is a need for longitudinal studies of patients before and after interventions to highlight off-target effects. For example, genetic studies of neuroticism and depression have shown enrichment of specific genetic downstream networks in targets of antidepressant drugs (Wary N R et al 2018 and Luciano M et al 2018). This is possible, as present therapies enrich known genetic networks. Such studies are possible in cohesive, large-scale collaborations involving data collections and biobanks across India for mental health programmes. In one such move, NIMHANS collaborated with National Centre for Biological Sciences (NCBS), Institute for Stem Cell and Regenerative Medicine (InSTEM), Christian Medical College (CMC) Vellore, and Department of Biotechnology (DBT), marking a giant effort in the right direction. The outcomes of such studies are much awaited.