However, integration – in other words, jointly analyzing data from multiple sources (also referred to as data fusion) - is challenging due to the different nature of these data sets. Dynamic data sets can often be arranged as a multiway array, e.g., participants by metabolites by time tensor, while static data sets can be in the form of matrices, e.g., participants by genes. Tensor factorizations, which are extensions of matrix factorizations (e.g., principal component analysis) to multiway data sets, have been successfully used to reveal the underlying patterns in such higher-order/multiway data sets. For instance, tensor factorizations have been used to analyze gut microbiome data collected from infants over time and capture birth mode-related microbial changes, have shown promising performance in terms of revealing temporal phenotypes from electronic health records as well as capturing metabolic differences among participants in terms of their response to a meal challenge test. Tensor factorizations have been extended to joint analysis of data sets from multiple sources through coupled matrix and tensor factorizations, and such coupled factorization-based approaches have also started to be used in systems biology applications.

In this workshop, we will bring together a selection of studies (where integration of such heterogeneous data is of interest) and the researchers working on developing methods based on tensor factorizations and their extensions to joint analysis of data from multiple sources with a particular interest in omics data analysis.  We expect the workshop to facilitate the synergy between different disciplines and pave the way to effective methods to address data analysis challenges, especially in terms of integration of data sets, in precision health.