[Surveillance and detection of unusual events in toxicovigilance: Review of relevant methods]. / Surveillance et détection des événements inhabituels en toxicovigilance : revue des méthodes pertinentes.

BACKGROUND: Declared cases of exposures related to potential toxic agents are reported through a national database, the French Network of Poison Centers, and account on average for 200,000 cases per year, including 75,000 to 80,000 symptomatic cases. These data are currently used to investigate signals from local, national or international institutional partners (such as hospitals, local health authorities, and the Rapid Alert System for Food and Feed). Our objective is to complete this classical toxicovigilance activity through the automated detection of unexpected or unusual events in order to identify precociously signals representing potential threats for public health. To reach this objective, the inventory of surveillance and detection methods of unexpected events is necessary. METHODS: A literature review was conducted via Scopus(®) and Pubmed(®) databases, completed with grey literature and data available on worldwide vigilance systems' websites. RESULTS: The most commonly used methods are disproportional measures in the field of pharmacovigilance, some of which are subject to a routine detection at regular time intervals. Criteria of signal generation differ from one system to another, which have implemented data filtering strategies before or after analysis, in order to decrease the number of generated signals and improve their priority level. These signals are then transmitted to an experts committee for a clinical and epidemiological evaluation, and at times, for informing the patient's medical records. We also notice an interest in other approaches such as surveillance methods of temporal series or symbolic methods for associative rules extraction between one or more drugs and one or more adverse effects, with the possibility to include other types of variables, such a demographic data. The developments of probabilistic-based algorithms have also been recently developed, opening new opportunities. CONCLUSION: These surveillance and detection methods are of high interest for the automated detection of signals from the French toxicovigilance network. The initial step to developing these methods consists in studying the statistical quality of data and targeting the needs and expectations of the toxicovigilance network for what we want and what we can detect.

Sequential patterns mining and gene sequence visualization to discover novelty from microarray data.

Data mining allow users to discover novelty in huge amounts of data. Frequent pattern methods have proved to be efficient, but the extracted patterns are often too numerous and thus difficult to analyze by end users. In this paper, we focus on sequential pattern mining and propose a new visualization system to help end users analyze the extracted knowledge and to highlight novelty according to databases of referenced biological documents. Our system is based on three visualization techniques: clouds, solar systems, and treemaps. We show that these techniques are very helpful for identifying associations and hierarchical relationships between patterns among related documents. Sequential patterns extracted from gene data using our system were successfully evaluated by two biology laboratories working on Alzheimer's disease and cancer.