Co-clustering of evolving count matrices with the dynamic latent block model: application to pharmacovigilance

Abstract

The simultaneous clustering of observations and features of datasets (known as co-clustering) has recently emerged as a central topic in machine learning applications. However, most models focus on continuous data in stationary scenarios, where cluster assignments do not evolve over time. We propose in this paper the dynamic latent block model (dLBM), which extends the classical binary latent block model, making amenable such analysis to dynamic cases where data are counts. Our approach operates on temporal count matrices allowing to detect abrupt changes in the way existing clusters interact with each other. The time breaks detection is performed through clustering of time instants, that allows for better model parsimony. The time dependent counting data are modeled via non-homogeneous Poisson processes (HHPPs), conditionally to the latent variables. In order to handle the model inference, we rely on a SEM-Gibbs algorithm and the ICL criterion is used for model selection. Numerical experiments on simulated data highlight the main features of the proposed approach and show the interest of dLBM with respect to related works. An application to adverse drug reaction in pharmacovigilance is also proposed, where dLBM was able to recognize clusters in a meaningful way that identified safety events that were consistent with retrospective knowledge. Hence, our aim is to propose this dynamic co-clustering method as a tool for automatic safety signal detection, to support medical authorities.

An artificial intelligence algorithm for co‐clustering to help in pharmacovigilance before and during the COVID‐19 pandemic

“…The data management and development of the co-clustering method were performed with R software version 4.1.2. 23 We used the timedependent co-clustering generative method, called dLBM, developed by Marchello et al 20 The data were organized so that the rows (drugs) were indexed by i = 1, …, N and the columns (ADRs) by j = 1, …, P.…”

“…Our dLBM approach in pharmacovigilance supports previous encouraging results. 20 Based on co-clustering, detection or strengthening of potential safety signals may emerge in an unsupervised manner. These approaches might pave the way for easier and more judicious analyses of massive upsurges of reports in pharmacovigilance databases, such as those observed since the launch of COVID-19 vaccines.…”

“…Previously, co-clustering using a dynamic latent block model (dLBM) has been used to 'unsupervisedly' detect drugs and ADRs clusters, according to appropriately chosen and equilibrated time periods ('clusters'). 20 Large datasets of drug-ADR combinations, evolving over time, could then be gathered and summarized into timedependent clusters. The use of dLBM had allowed the retrospective 'detection' of the main health crises that occurred in France during the period 2010-2020 (involving the drugs benfluorex, levonorgestrel and levothyroxine).…”

“…For each type of data (ADR, drug and period), co‐clustering brings together groups of similarly reacting items into a cluster. Previously, co‐clustering using a dynamic latent block model (dLBM) has been used to ‘unsupervisedly’ detect drugs and ADRs clusters, according to appropriately chosen and equilibrated time periods (‘clusters’) 20 . Large datasets of drug–ADR combinations, evolving over time, could then be gathered and summarized into time‐dependent clusters.…”

“…DISCUSSIONOur study further explored the relevance of unsupervised coclustering-based AI,20 when applied to pharmacovigilance. By means of a dLBM on a routine pharmacovigilance dataset constituted between 2012 and 2022, we have detected particular events of interest.…”