cluster_features: Cluster correlated features
In antonvsdata/amp: Smoothing Metabolomics Analyses

Description Usage Arguments Value See Also Examples

Clusters features potentially originating from the same compound. Features with high Pearson correlation coefficient and small retention time difference are linked together. Then clusters are formed by setting a threshold for the relative degree that each node in a cluster needs to fulfil. Each cluster is named after the feature with the highest median peak area (median abundance) This is a wrapper around numerous functions that are based on the MATLAB code by David Broadhurst.

1
2
3

cluster_features(object, all_features = FALSE, rt_window = 1/60,
  corr_thresh = 0.9, d_thresh = 0.8, mz_col = NULL, rt_col = NULL,
  plotting = FALSE, min_size_plotting = 3, prefix = NULL)

`object`	a MetaboSet object
`all_features`	logical, should all features be included in the clustering? If FALSE as the default, flagged features are not included in clustering
`rt_window`	the retention time window for potential links NOTE: use the same unit as the retention time
`corr_thresh`	the correlation threshold required for potential links between features
`d_thresh`	the threshold for the relative degree required by each node
`mz_col`	the column name in fData(object) that holds mass-to-charge ratios
`rt_col`	the column name in fData(object) that holds retention times
`plotting`	should plots be drawn for each cluster?
`min_size_plotting`	the minimum number of features a cluster needs to have to be plotted
`prefix`	the prefix to the files to be plotted

a MetaboSet object, with median peak area (MPA), the cluster ID, the features in the cluster, and cluster size added to results.

find_connections, find_clusters, visualize_clusters, assign_cluster_id, compress_clusters

1 2	# The parameters are really weird because example data is imaginary clustered <- cluster_features(example_set, rt_window = 1, corr_thresh = 0.5, d_thresh = 0.6)