The translation of precision medicine in clinical practice will depend mostly from the possibility to make statistical inference at individual level, exactly positioning a new case in the taxonomy space (diagnosis) or in the time space (prognosis). As president Barak Obama said in 2015 in occasion of the launch of Precision Medicine Initiative “Most medical treatments have been designed for the “average patient”. We know otherwise that every patient is a person with unique characteristics. An intervention may be effective for a population but not necessarily for that individual patient. The recommendation of a guideline may not be right for a particular patient because it is not what he or she wants, and implementing the recommendation will not necessarily mean a favorable outcome. As matter of the fact clinical epidemiology and medical statistics have not been suited to answer specific questions at the individual level. They focus on groups of individuals and not on single individuals. Classical statistics by definition needs samples to work, and samples by definition are always greater than one. This explains why for traditional statistics the single individual is a sort of moving and vague target to intercept. Statistical predictive models can fail dramatically when applied to the single individual. In a model that has an overall 90% accuracy in predicting an event on a group level, the degree of confidence can drop substantially when applied to a single subject.One of the major challenges to delivering effective treatment is to devise a method capable of determining the confidence interval of a single individual. This is now possible by feeding the patient data into machine learning systems which are able to map the patient as an “hyperpoint” in a multidimensional space where the patient variables are the coordinates and to measure his “distance” or more simply his similarity with other patients of whom the target of interest ( diagnosis, prognosis or response to a certain treatment) is already known. The neigh borough statistics of this new patient therefore could allow us to calculate the confidence interval pertaining to him, and therefore to make inference at individual level. In this paper we describe how Machine learning systems open new avenues to this formidable task allowing to approach in a consistent and sound way the problem of single individual statistics. As exemplification we propose four complementary examples and approaches using unsupervised (first 3 examples) and supervised (fourth example) machine learning systems.

This paper, entitled "Machine Learning Systems and Precision Medicine: a Conceptual and Experimental Approach to Single Individual Statistics" is an original study on the role of machine learning system in allowing to approach in a consistent and sound way the problem of single individual statistics, a fundamental scope of Precision Medicine Initiative. In the last years there has been an explosion of papers published on precision medicine topic, but very few of them, if any have answered to this very simple question: how it is possible to translate with high accuracy the group statistics at an individual level in term of confidence interval. The experience gained with the use of special machine learning systems developed at Semeion research Centre has allowed us to describe three cases studies relevant to precision medicine approach with different unsupervised machine learning systems. The three methods proved to be reliable and easily applicable to real world examples in term of readability, accuracy and reproducibility and seem to have the potential to allow the real translation of precision medicine philosophy in the real world. In a future paper we will cover the topic with supervised machine learning system, through the Fermi mathematics

Aims: This study examined the prevalence of tobacco, alcohol, and cannabis use and their correlates among Italian High School students. Methods: A nationwide online epidemiological survey based on the European School Survey Project on Alcohol and Other Drugs was administered to Italian high school students in 2012 using two-stage probabilistic cluster sampling. There were 35,980 student respondents with analyzable data drawn from 490 schools. The mean age of respondents was 17.0 (SD=1.4) years and 50.0% were female. The association between past-30-day cigarette, alcohol, and drug use and 11 correlates identified in the literature as potentially important in understanding such use was examined in three separate logistic regression analyses. A critical value of 363.4 was determined for the Wald χ2 test that allowed detection of a small effect size. Results: Lifetime cigarette, alcohol, and cannabis use was reported by 22,259 (61.9%), 30,754 (85.5%), and 8.142 (22.6%) of the sample, respectively. Past-30-day cigarette smoking was significantly associated with going out a lot at night (AOR=2.2), more than one day skipping school in the past 30 days (AOR=1.7), lifetime alcohol intoxication (AOR=2.9), lifetime use of at least one drug (AOR=8.4), and friends/siblings who use drugs (AOR=1.8). Past-30-day alcohol use was significantly associated with going out a lot at night (AOR=2.8), lifetime cigarette use (AOR=2.4), use of at least one drug lifetime (AOR=2.3), friends and siblings who use drugs (AOR=1.9), and male gender (AOR=1.6). Significant correlates of past-30-day cannabis use were lifetime cigarette use (AOR=13.7), lifetime alcohol intoxication (AOR=4.0), and having friends/siblings who use illicit drugs (AOR=21.4). Conclusions: Tobacco, alcohol, and cannabis use is prevalent among Italian high school students. Pediatricians should screen patients for their use. Additional research is needed to find effective approaches to delay and reduce substance use among youth that addresses the significant correlates of use.