Multi-view clustering refers to clustering samples that have multiple data representations. This is the case for multi-omics data measured on the same samples, where each data type is a separate representation of the same samples. All multi-view methods supported by OmicsAnalyst have the same general approach, which involves:
- computing a sample similarity matrix from each ‘omics type individually;
- integrating the individual matrices together;
- determining the optimal number of sample clusters in the integrated matrix; and
- detecting this number of clusters.
A main advantage of multi-view clustering is that it tends to reduce random noise or platform-specific technical artifacts, as it is highly unlikely that exact same erroneous effects are present across multiple data sets. For more details, please refer to this excellent review.