Unsupervised Analysis of Classical Biomedical Markers: Robustness and Medical Relevance of Patient Clustering Using Bioinformatics Tools

Abstract

MotivationIt has been proposed that clustering clinical markers, such as blood test results, can be used to stratify patients. However, the robustness of clusters formed with this approach to data pre-processing and clustering algorithm choices has not been evaluated, nor has clustering reproducibility. Here, we made use of the NHANES survey to compare clusters generated with various combinations of pre-processing and clustering algorithms, and tested their reproducibility in two separate samples.MethodValues of 44 biomarkers and 19 health/life style traits were extracted from the National Health and Nutrition Examination Survey (NHANES). The 1999–2002 survey was used for training, while data from the 2003–2006 survey was tested as a validation set. Twelve combinations of pre-processing and clustering algorithms were applied to the training set. The quality of the resulting clusters was evaluated both by considering their properties and by comparative enrichment analysis. Cluster assignments were projected to the validation set (using an artificial neural network) and enrichment in health/life style traits in the resulting clusters was compared to the clusters generated from the original training set.ResultsThe clusters obtained with different pre-processing and clustering combinations differed both in terms of cluster quality measures and in terms of reproducibility of enrichment with health/life style properties. Z-score normalization, for example, dramatically improved cluster quality and enrichments, as compared to unprocessed data, regardless of the clustering algorithm used. Clustering diabetes patients revealed a group of patients enriched with retinopathies. This could indicate that routine laboratory tests can be used to detect patients suffering from complications of diabetes, although other explanations for this observation should also be considered.ConclusionsClustering according to classical clinical biomarkers is a robust process, which may help in patient stratification. However, optimization of the pre-processing and clustering process may be still required.

Cardiometabolic and renal phenotypes and transitions in the United States population

“…Finally, our results on ethnic and educational disparities in the prevalence of specific clusters were consistent with previous studies that considered risk factors either individually 36,78 or through the lens of optimal cardiometabolic health 23 , but these studies did not examine disparities in a comprehensive set of cardiometabolic and renal phenotypes of risk factors. Our results are not directly comparable with those using electronic health records due to differences in the study population, methods and clinical conditions used in the clustering and because some of these studies aimed at identifying subtypes of specific diseases 45,47,48,50,51,[53][54][55][56][57][58][59][60][61][62][63][64][65][66][67][68] . Among such studies, two studies in different populations identified phenotypes characterized by compromised kidney function and low DBP 50,56 .…”

The Identification of Diabetes Mellitus Subtypes Applying Cluster Analysis Techniques: A Systematic Review

“…Second, the type of data (observational/longitudinal) is also critical in cluster analysis to give a chance to observe temporal patterns of disease progression, as cluster analysis does not explain the aetiology of the disease. Third, the number of clusters depends on the specific methodology applied as well as the proportions of populations among clusters that could vary based on the chosen sample size and the presence/absence of scaling the dataset (preprocessing) [ 25 ].…”

Identification of Mitophagy-Associated Genes for the Prediction of Metabolic Dysfunction-Associated Steatohepatitis Based on Interpretable Machine Learning Models

“…An unsupervised clustering analysis was performed using the “ConsensusClusterPlus” R package. 18 The k-means approach was run 1000 times with a maximum subtype number of k (k = 6). The optimal subtype number was ascertained using the cumulative distribution function curve, consensus matrix, and consistent cluster score.…”