Using Artificial Intelligence to diagnose and detect Asthma sub-types in children
06 Feb 2018
By Dr. Tavpritesh Sethi, Early Career Fellow
All India Institute of Medical Sciences, New Delhi
Advances in genomics and molecular sciences are helping us refine the disease definitions and to tailor treatments. Such ‘precision-medicine’ or ‘stratified-medicine’ strategies often rely upon discovery of new endotypes, i.e. sub-divisions of a disease. In diseases such as breast cancer, molecular stratification helps doctors to choose drugs and save more lives. In complex diseases such as asthma and diabetes also, clinical experience indicates the existence of endotypes. However, these are much harder to discover because common diseases are also more nuanced (complex).
We addressed the challenge of discovering endotypes of childhood asthma by analyzing breath through chemical patterns which are difficult to see by human eye. Since artificial-intelligence and machine-learning have been shown to perform better than humans in complex-repetitive tasks, we used these for pattern discovery.
The study was conducted at All India Institute of Medical Sciences (AIIMS), New Delhi, where children were monitored over a period of five years and the experiments were conducted at CSIR-Institute of Genomics and Integrative Biology., New Delhi At the time of first enrollment, the children breathed into an apparatus that condensed the water-vapor and metabolites in their breath (Exhaled Breath Condensate, EBC), a reflection of their lungs and airway chemistry. EBC was then analyzed by Nuclear Magnetic Resonance (NMR) spectroscopy, a technique commonly used to understand the chemical composition of mixtures. The NMR spectra generated were used to train a machine-learning algorithms (Random Forest) which works like a committee of experts, evaluating a candidate as asthmatic or non-asthmatic and further dividing asthma into endotypes. These endotypes were learnt by the computer without expert input and served as a snapshot of airway chemistry at enrollment time.
While these endotypes of asthmatic children did not differ in severity of disease at enrollment, they had markedly different 5-year incidence of breakthrough exacerbations despite being on treatment. This not only expands our understanding breath-chemistry associated with long-term exacerbations in asthma, but also suggests the potential of deciding treatment strategies tailored to the discovered endotypes to improve outcomes in childhood asthma.
Exhaled breath condensate metabolome clusters for endotype discovery in asthma. Anirban Sinha, Koundinya Desiraju*, Kunal Aggarwal*, Rintu Kutum, Siddhartha Roy, Rakesh Lodha, S. K. Kabra, Balaram Ghosh, Tavpritesh Sethi and Anurag Agrawal
Banner image credit - Dave Farnham. Wellcome Images. Internal architecture of the trachea and lungs, 3D printed in Frosted Ultra Detail plastic.