Identification of disease-distinct complex biomarker patterns by means of unsupervised machine-learning using an interactive R toolbox (Umatrix)

Abstract

Background: Unsupervised machine-learned analysis of cluster structures, applied using the emergent self-organizing feature maps (ESOM) combined with the unified distance matrix (U-matrix) has been shown to provide an unbiased method to identify true clusters. It outperforms classical hierarchical clustering algorithms that carry a considerable tendency to produce erroneous results. To facilitate the application of the ESOM/U-matrix method in biomedical research, we introduce the interactive R-based bioinformatics tool "Umatrix", which enables valid identification of a biologically meaningful cluster structure in the data by training a Kohonen-type self-organizing map followed by interface-guided interactive clustering on the emergent U-matrix map. Results: The ability to detect clinical relevant subgroups was applied to a data set comprising plasma concentrations of d = 25 lipid markers including endocannabinoids, lysophosphatidic acids, ceramides and sphingolipids acquired from n = 100 patients with Parkinson's disease and n = 100 controls. Following ESOM training, clear data structures in the high-dimensional data space were observed on the U-matrix, allowing separation of patients from controls almost perfectly. When the data structure was destroyed by Monte-Carlo random resampling, the U-matrix became unstructured and patients and controls were mixed. Obtained results are biologically plausible and supported by empirical evidence of a regulation of several classes of lipids in Parkinson's disease. Conclusions: Sophisticated analysis of structures in biomedical data provides a basis for the mechanistic interpretation of the observations and facilitates subsequent analyses focusing on hypothesis testing. The freely available R library "Umatrix" provides an interactive tool for broader application of unsupervised machine learning on complex biomedical data.

A machine‐learned analysis of human gene polymorphisms modulating persisting pain points to major roles of neuroimmune processes

“…Valleys indicate clusters of functionally similar genes based on the significant GO term annotations. The figure was created using the R software package (version 3.3.2 for Linux; http://CRAN.R-project.org/; R Development Core Team, ) using our R library ‘Umatrix’ ( https://cran.r-project.org/package=Umatrix; Lötsch et al., ).…”

“…The ‘map’ was further enhanced by calculating a so‐called P‐matrix (Ultsch, ), which displays the probability of a data point as estimated as the number of data points in a sphere with radius r around x at each grid point on the ESOM's output grid. The calculations were performed using our R library ‘Umatrix’ (https://cran.r-project.org/package=Umatrix; Lötsch et al., ).…”

Current Projection Methods-Induced Biases at Subgroup Detection for Machine-Learning Based Data-Analysis of Biomedical Data

“…When analyzing the same data using emergent self-organizing feature maps (ESOM), projecting the data on a grid of thousands of artificial neurons [7] and combining it with U-matrix methods [4,5], it was correctly concluded that there is no group structure in these data. As shown in the lower-left panel of Figure 1, the U-matrix consists of a random structure without subgroups.…”

“…Results of an alternative projection and subgroup detection technique, implemented as ESOM/U-matrix, which clearly indicate the absence of any systematic data structures, are shown at lower-left panel. The figure has been created based on the t-SNE analysis implemented in the R library "tsne" [3] and the U-matrix was obtained using the R library "Umatrix" [4].…”

“…With an emphasis on t-SNE as a data projection technique that appears to be becoming a standard in flow cytometry, this paper, therefore, aims to evaluate the results of this technique when applied on several different artificial or biomedical data sets. To demonstrate that data analysis is not limited to a single method that is occasionally delivered with the laboratory device, an alternative approach has been used in parallel to t-SNE, consisting of emergent self-organizing feature maps (ESOM) that have been shown to reliably detect structures in artificial and biomedical data sets [4,5]. The following report will identify potential pitfalls in the use of t-SNE for the analysis of data in molecular research and provide some technical background to these shortcomings; however, it cannot substitute the fundamentals of data science and tuning of analytical methods, which requires further reading presented in a broader context such as [6].…”