Global Considerations in Hierarchical Clustering Reveal Meaningful Patterns in Data

Abstract

BackgroundA hierarchy, characterized by tree-like relationships, is a natural method of organizing data in various domains. When considering an unsupervised machine learning routine, such as clustering, a bottom-up hierarchical (BU, agglomerative) algorithm is used as a default and is often the only method applied.Methodology/Principal FindingsWe show that hierarchical clustering that involve global considerations, such as top-down (TD, divisive), or glocal (global-local) algorithms are better suited to reveal meaningful patterns in the data. This is demonstrated, by testing the correspondence between the results of several algorithms (TD, glocal and BU) and the correct annotations provided by experts. The correspondence was tested in multiple domains including gene expression experiments, stock trade records and functional protein families. The performance of each of the algorithms is evaluated by statistical criteria that are assigned to clusters (nodes of the hierarchy tree) based on expert-labeled data. Whereas TD algorithms perform better on global patterns, BU algorithms perform well and are advantageous when finer granularity of the data is sought. In addition, a novel TD algorithm that is based on genuine density of the data points is presented and is shown to outperform other divisive and agglomerative methods. Application of the algorithm to more than 500 protein sequences belonging to ion-channels illustrates the potential of the method for inferring overlooked functional annotations. ClustTree, a graphical Matlab toolbox for applying various hierarchical clustering algorithms and testing their quality is made available.ConclusionsAlthough currently rarely used, global approaches, in particular, TD or glocal algorithms, should be considered in the exploratory process of clustering. In general, applying unsupervised clustering methods can leverage the quality of manually-created mapping of proteins families. As demonstrated, it can also provide insights in erroneous and missed annotations.

Dissecting Cooperative Communications in a Protein with a High‐Throughput Single‐Molecule Scalpel

“…A similar deconvolution approach previously yielded a resolution of 0.5 nm. [51] We performed a two-stage HCA. [51] We performed a two-stage HCA.…”

“…[17] HCA Matlab programs were used to perform HCA, according to the literature. [51] We performed a two-stage HCA. In the first stage, we evaluated the distance between the two molecules (i and j) according to the formula ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi…”

Neurocircuitry for modeling drug effects

“…Organizing complex datasets into hierarchies is a common strategy in many fields and hierarchical clustering has been successfully applied to many neurobiologically oriented questions. It is assumed that bottom‐up (agglomerative) hierarchical algorithms have some advantages in identifying local relations in the data, whereas top‐down methods, such as divisive hierarchical algorithms, better capture global patterns (Varshavsky, Horn & Linial ). By using an assumption‐free data‐mining approach with divisive hierarchical algorithms, we identified brain sites of potential importance in mediating alcohol and drug effects that are likely to be relevant to drug reinforcement, and then applied agglomerative clustering to obtain a functional neurochemical neurocircuitry that provides a comprehensive structural framework for large‐scale mathematical modeling.…”

Discovery and characterisation of dietary patterns in two Nordic countries

“…[82][83][84] Thus, entities within an accordingly identified cluster are more similar, relative to those appearing in other aggregations. Such clusters may e.g.…”

“…Algorithms for divisive (top-down) hierarchical clustering (DHC), on the other hand, which iteratively split all the N samples into a hierarchy of subclusters, are more computation-intensive but typically create more reliable groups at higher branching levels. [83][84][85] In analogy with standard agglomerative clustering, most DHC methods still create binary (or alternatively a fixed predefined value of) divisions at each branching point. Flexible DHC algorithms allowing multiple branches at each branching level have, however, greater potential to reveal more meaningful groupings that are also easier to interpret.…”