کلان داده در علوم پزشکی و پزشکی قانونی Big data in forensic science and medicine

Introduction In less than a decade, big data in medicine has become quite a phenomenon and many biomedical disciplines got their own tribune on the topic. Perspectives and debates are flourishing while there is a lack for a consensual definition for big data. Big data presents all the attributes of a buzz word, but it should be thoroughly investigated so that it can be decided whether big data is a mere trend or the premises of a true revolution for research and routine practice. Origins of the term “big data” are unclear and we propose here to go many years back in time to search for them. Techniques usually presented as specific to big data such as machine learning techniques are supposed to support the ambition of personalized, predictive and preventive medicines. These techniques are mostly far from being new and are more than 50 years old for the most ancient. On the other hand, several issues closely related to the properties of big data and inherited from other scientific fields such as artificial intelligence are often underestimated if not ignored. In this paper, we expose the most important of these issues. Finally, in a context of general enthusiasm about the big data phenomenon, forensic science is still awaiting for their position papers as well as for a comprehensive outline of what kind of contribution big data could bring to the field. The present situation calls for definitions and actions to rationally guide research and practice in big data. Here, we briefly present what is at stakes for forensic science if it wants to embrace the philosophy of the big data era while avoiding its main pitfalls. 1 – Is big data more than a buzz word and where does it come from? No one can determine where or when the use of the term “big data” originated (1). It exponentially spread and contaminated all scientific and non-scientific fields within the past decade. 1.1 Origins of big data: definitions and practices of big data in the past decade. Big data is a vague and generic term that can encompass several distinct and non-exclusive properties. The origin of the big data concept is often attributed to a short technical report from the META group which is an American consulting firm, since become Gartner (2). This report was written by Doug Laney in 2001 and presented the challenge of the “3D Data management” which evolved into the 3Vs concept: Volume – Variety – Velocity. Other Vs have been proposed since then, such as Veracity. All Vs refer to data: the challenges that faces the information society are bound to great Volumes of Various and heterogeneous data to process in real-time (Velocity). Practically, there is no consensual definition of big data. Even if many recognized the 3Vs terminology, not all understand the same meaning for each V. Volume may be the most cited property of big data, maybe because echoing to the “big” part. What is Volume? If data are figured by a 2-dimensional table (columns for variables and lines for observations or patients) is a big Volume of data a table with many observations (lines) or with many variables (columns)? Or both? Baro and colleagues also suggested defining Volume as a combination of both these dimensions (3) (figure1). The distinction seems theoretic but has strong implications, since it is the number of variables that defines the dimensionality of data, and not the number of observations (see part 2 – the curse of dimensionality).