Limits...
Unsupervised analysis of classical biomedical markers: robustness and medical relevance of patient clustering using bioinformatics tools.

Markovich Gordon M, Moser AM, Rubin E - PLoS ONE (2012)

Bottom Line: It has been proposed that clustering clinical markers, such as blood test results, can be used to stratify patients.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.However, optimization of the pre-processing and clustering process may be still required.

View Article: PubMed Central - PubMed

Affiliation: Sharaga Segal Dept of Microbiology and Immunology, Ben Gurion University of the Negev, Beersheba, Israel.

ABSTRACT

Motivation: It 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.

Method: Values 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.

Results: The 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.

Conclusions: Clustering 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.

Show MeSH

Related in: MedlinePlus

Selected clusters from the NHANES diabetic subset.(A) The mean and standard deviation of biomarker values are shown for four selected clusters generated with 299 males 20 years of age or older from the NHANES training set with self-reported type 2 diabetes mellitus. Clustering was performed using the Z-score normalized with age adjustment/CLICK pre-processing/clustering combination. For each cluster, the total number of individuals (top, right) and selected health/lifestyle traits that are significantly enriched in that cluster (top, left) are provided. For each enriched term, the enrichment factor (i.e. the frequency of the term in a cluster divided by its frequency in the entire dataset) is also provided. (B) Comparison of the glucose levels in clusters 1, 2, 5 and 6.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3293863&req=5

pone-0029578-g003: Selected clusters from the NHANES diabetic subset.(A) The mean and standard deviation of biomarker values are shown for four selected clusters generated with 299 males 20 years of age or older from the NHANES training set with self-reported type 2 diabetes mellitus. Clustering was performed using the Z-score normalized with age adjustment/CLICK pre-processing/clustering combination. For each cluster, the total number of individuals (top, right) and selected health/lifestyle traits that are significantly enriched in that cluster (top, left) are provided. For each enriched term, the enrichment factor (i.e. the frequency of the term in a cluster divided by its frequency in the entire dataset) is also provided. (B) Comparison of the glucose levels in clusters 1, 2, 5 and 6.

Mentions: To further expand and test this hypothesis, we performed unsupervised clustering to detect sub-populations of diabetic individuals. A subset of diabetic individuals (i.e. males, 20 years of age or more) was extracted from the training and the validation sets described above (N = 299 for the training dataset and N = 229 for the validation dataset) by adding two biomarkers that are routinely tested in diabetes patients, namely BMI (body mass index) and glycosylated hemoglobin levels (A1C). The resulting dataset was clustered using the same procedure as described above (Figure 1) but this time using only the CLICK/Z-normalization with age adjustment pipeline. This procedure yielded six clusters involving 245 individuals, with 54 individuals remaining unclassified. To avoid confusion with the clusters discussed above, these clusters are named D1 through D6. The resulting clusters differ from each other in terms of their mean patterns (Figure 3). Cluster D2 members, for instance, are characterized by only slightly elevated glucose levels (with a mean of 111.9±35 mg/dL), cluster D1 members are characterized by moderately increased mean glucose levels (mean of 148.9±56 mg/dL), while individuals in cluster D5 have very high glucose levels (mean of 277.4±108 mg/dL).


Unsupervised analysis of classical biomedical markers: robustness and medical relevance of patient clustering using bioinformatics tools.

Markovich Gordon M, Moser AM, Rubin E - PLoS ONE (2012)

Selected clusters from the NHANES diabetic subset.(A) The mean and standard deviation of biomarker values are shown for four selected clusters generated with 299 males 20 years of age or older from the NHANES training set with self-reported type 2 diabetes mellitus. Clustering was performed using the Z-score normalized with age adjustment/CLICK pre-processing/clustering combination. For each cluster, the total number of individuals (top, right) and selected health/lifestyle traits that are significantly enriched in that cluster (top, left) are provided. For each enriched term, the enrichment factor (i.e. the frequency of the term in a cluster divided by its frequency in the entire dataset) is also provided. (B) Comparison of the glucose levels in clusters 1, 2, 5 and 6.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3293863&req=5

pone-0029578-g003: Selected clusters from the NHANES diabetic subset.(A) The mean and standard deviation of biomarker values are shown for four selected clusters generated with 299 males 20 years of age or older from the NHANES training set with self-reported type 2 diabetes mellitus. Clustering was performed using the Z-score normalized with age adjustment/CLICK pre-processing/clustering combination. For each cluster, the total number of individuals (top, right) and selected health/lifestyle traits that are significantly enriched in that cluster (top, left) are provided. For each enriched term, the enrichment factor (i.e. the frequency of the term in a cluster divided by its frequency in the entire dataset) is also provided. (B) Comparison of the glucose levels in clusters 1, 2, 5 and 6.
Mentions: To further expand and test this hypothesis, we performed unsupervised clustering to detect sub-populations of diabetic individuals. A subset of diabetic individuals (i.e. males, 20 years of age or more) was extracted from the training and the validation sets described above (N = 299 for the training dataset and N = 229 for the validation dataset) by adding two biomarkers that are routinely tested in diabetes patients, namely BMI (body mass index) and glycosylated hemoglobin levels (A1C). The resulting dataset was clustered using the same procedure as described above (Figure 1) but this time using only the CLICK/Z-normalization with age adjustment pipeline. This procedure yielded six clusters involving 245 individuals, with 54 individuals remaining unclassified. To avoid confusion with the clusters discussed above, these clusters are named D1 through D6. The resulting clusters differ from each other in terms of their mean patterns (Figure 3). Cluster D2 members, for instance, are characterized by only slightly elevated glucose levels (with a mean of 111.9±35 mg/dL), cluster D1 members are characterized by moderately increased mean glucose levels (mean of 148.9±56 mg/dL), while individuals in cluster D5 have very high glucose levels (mean of 277.4±108 mg/dL).

Bottom Line: It has been proposed that clustering clinical markers, such as blood test results, can be used to stratify patients.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.However, optimization of the pre-processing and clustering process may be still required.

View Article: PubMed Central - PubMed

Affiliation: Sharaga Segal Dept of Microbiology and Immunology, Ben Gurion University of the Negev, Beersheba, Israel.

ABSTRACT

Motivation: It 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.

Method: Values 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.

Results: The 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.

Conclusions: Clustering 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.

Show MeSH
Related in: MedlinePlus