Three dissimilarity measures to contrast dendrograms

How frequently do clusters occur in hierarchical clustering analysis? A graph theoretical approach to studying ties in proximity

Background

Hierarchical cluster analysis (HCA) is a widely used classificatory technique in many areas of scientific knowledge. Applications usually yield a dendrogram from an HCA run over a given data set, using a grouping algorithm and a similarity measure. However, even when such parameters are fixed, ties in proximity (i.e. two equidistant clusters from a third one) may produce several different dendrograms, having different possible clustering patterns (different classifications). This situation is usually disregarded and conclusions are based on a single result, leading to questions concerning the permanence of clusters in all the resulting dendrograms; this happens, for example, when using HCA for grouping molecular descriptors to select that less similar ones in QSAR studies.

Results

Representing dendrograms in graph theoretical terms allowed us to introduce four measures of cluster frequency in a canonical way, and use them to calculate cluster frequencies over the set of all possible dendrograms, taking all ties in proximity into account. A toy example of well separated clusters was used, as well as a set of 1666 molecular descriptors calculated for a group of molecules having hepatotoxic activity to show how our functions may be used for studying the effect of ties in HCA analysis. Such functions were not restricted to the tie case; the possibility of using them to derive cluster stability measurements on arbitrary sets of dendrograms having the same leaves is discussed, e.g. dendrograms from variations of HCA parameters. It was found that ties occurred frequently, some yielding tens of thousands of dendrograms, even for small data sets.

Conclusions

Our approach was able to detect trends in clustering patterns by offering a simple way of measuring their frequency, which is often very low. This would imply, that inferences and models based on descriptor classifications (e.g. QSAR) are likely to be biased, thereby requiring an assessment of their reliability. Moreover, any classification of molecular descriptors is likely to be far from unique. Our results highlight the need for evaluating the effect of ties on clustering patterns before classification results can be used accurately.Graphical abstract

A hierarchical method for whole-brain connectivity-based parcellation

In modern neuroscience there is general agreement that brain function relies on networks and that connectivity is therefore of paramount importance for brain function. Accordingly, the delineation of functional brain areas on the basis of diffusion magnetic resonance imaging (dMRI) and tractography may lead to highly relevant brain maps. Existing methods typically aim to find a predefined number of areas and/or are limited to small regions of grey matter. However, it is in general not likely that a single parcellation dividing the brain into a finite number of areas is an adequate representation of the function-anatomical organization of the brain. In this work, we propose hierarchical clustering as a solution to overcome these limitations and achieve whole-brain parcellation. We demonstrate that this method encodes the information of the underlying structure at all granularity levels in a hierarchical tree or dendrogram. We develop an optimal tree building and processing pipeline that reduces the complexity of the tree with minimal information loss. We show how these trees can be used to compare the similarity structure of different subjects or recordings and how to extract parcellations from them. Our novel approach yields a more exhaustive representation of the real underlying structure and successfully tackles the challenge of whole-brain parcellation.

Cophenetic metrics for phylogenetic trees, after Sokal and Rohlf

BACKGROUND

Phylogenetic tree comparison metrics are an important tool in the study of evolution, and hence the definition of such metrics is an interesting problem in phylogenetics. In a paper in Taxon fifty years ago, Sokal and Rohlf proposed to measure quantitatively the difference between a pair of phylogenetic trees by first encoding them by means of their half-matrices of cophenetic values, and then comparing these matrices. This idea has been used several times since then to define dissimilarity measures between phylogenetic trees but, to our knowledge, no proper metric on weighted phylogenetic trees with nested taxa based on this idea has been formally defined and studied yet. Actually, the cophenetic values of pairs of different taxa alone are not enough to single out phylogenetic trees with weighted arcs or nested taxa.

RESULTS

For every (rooted) phylogenetic tree T, let its cophenetic vectorφ(T) consist of all pairs of cophenetic values between pairs of taxa in T and all depths of taxa in T. It turns out that these cophenetic vectors single out weighted phylogenetic trees with nested taxa. We then define a family of cophenetic metrics dφ,p by comparing these cophenetic vectors by means of Lp norms, and we study, either analytically or numerically, some of their basic properties: neighbors, diameter, distribution, and their rank correlation with each other and with other metrics.

CONCLUSIONS

The cophenetic metrics can be safely used on weighted phylogenetic trees with nested taxa and no restriction on degrees, and they can be computed in O(n2) time, where n stands for the number of taxa. The metrics dφ,1 and dφ,2 have positive skewed distributions, and they show a low rank correlation with the Robinson-Foulds metric and the nodal metrics, and a very high correlation with each other and with the splitted nodal metrics. The diameter of dφ,p, for p⩾1 , is in O(n(p+2)/p), and thus for low p they are more discriminative, having a wider range of values.

A polynomial-time algorithm computing lower and upper bounds of the rooted subtree prune and regraft distance

Rooted, leaf-labeled trees are used in biology to represent hierarchical relationships of various entities, most notably the evolutionary history of molecules and organisms. Rooted Subtree Prune and Regraft (rSPR) operation is a tree rearrangement operation that is used to transform a tree into another tree that has the same set of leaf labels. The minimum number of rSPR operations that transform one tree into another is denoted by d(rSPR) and gives a measure of dissimilarity between the trees, which can be used to compare trees obtained by different approaches, or, in the context of phylogenetic analysis, to detect horizontal gene transfer events by finding incongruences between trees of different evolving characters. The problem of computing the exact d(rSPR) measure is NP-hard, and most algorithms resort to finding sequences of rSPR operations that are sufficient for transforming one tree into another, thereby giving upper bound heuristics for the distance. In this article, we present an O(n⁴) recursive algorithm D-Clust that gives both lower bound and upper bound heuristics for the distance between trees with n shared leaves and also gives a sequence of operations that transforms one tree into another. Our experiments on simulated pairs of trees containing up to 100 leaves showed that the two bounds are almost equal for small distances, thereby giving the nearly-precise actual value, and that the upper bound tends to be close to the upper bounds given by other approaches for all pairs of trees.