Non-negative matrix factorisation (NMF) is an unsupervised machine learning technique for identifying patterns in data. NMF has a long history of use in bioinformatics and a number of NMF methods have been developed specifically for omics data. This also includes semi-supervised approaches to NMF.
However, how to best apply unsupervised or semi-supervised approaches is currently not clear. This project will deliver a benchmark of NMF methods, and develop new methods and guidelines to make these tools more accessible and their results more interpretable.