HeatmapR - Create Heatmaps Using Base Graphics
Provides a lightweight framework for creating high
quality, complex heatmaps using base graphics. Supports
hierarchical clustering with dendrograms, column and row
scaling, cluster sub-divisions, customisable cell colours,
shapes and sizes, legends, and flexible layouts for arranging
multiple heatmaps. Designed to return plot objects that can be
easily arranged with other plots without sacrificing
resolution. Methods for hierarchical clustering and distance
computations are described in Murtagh and Contreras (2012)
<doi:10.1002/widm.53>; linkage uses the nearest-neighbour chain
algorithm, a minimum spanning tree for single linkage and the
generic algorithm of Mullner (2011)
<doi:10.48550/arXiv.1109.2378> for centroid and median linkage,
leaves are ordered optimally as in Bar-Joseph, Gifford and
Jaakkola (2001) <doi:10.1093/bioinformatics/17.suppl_1.S22>,
Kendall correlations are computed by Knight's (1966) algorithm
<doi:10.1080/01621459.1966.10480879>, k-means by Hartigan and
Wong (1979) <doi:10.2307/2346830>, also from the k-means++
starts of Arthur and Vassilvitskii (2007), and the dip
statistic of unimodality, which flags groups of cells that may
be two populations, by Hartigan and Hartigan (1985)
<doi:10.1214/aos/1176346577>, all in compiled code.