Use of similarity metrics aiming to identify groupings of entities with clinical and biological significance

Precision medicine is an emerging concept of medicine which aims at refining the perception of disease and treatment in order to obtain better patient outcomes. A potent tool in the development of precision medicine is that of similarity networks which aim to identify relations between relevant items based on similarity and visualised as networks of association. We wanted to assess the effectiveness of such tool in using different types of data in different fields of medical enquiry with clinical relevance. The three main domains in which we applied these methods were at the level of the cell, patient and disease, attempting to find meaningful relationships between elements of these three domains. With respect to cells, we aimed to develop a similarity matching network in order to identify cell types. With respect to patients we aimed to develop a collection of similarity networks aimed at identifying endotypes (subgroups of patient with clinical relevance). With respect to diseases we aimed to develop a similarity network to identify relationships between diseases. In each of the three domains we developed such networks and compared them to currently used alternatives in order to assess their effectiveness at performing their desired goals and uncovering interesting groups and relations whether they are groups of patients, types of cells or diseases. Our application showed that such techniques can have outputs comparable to current alternatives and have the potential to be used meaningfully in discovering such relations as to facilitate the advancement of the objectives of precision medicine