Assessment of Drugs Toxicity and Associated Biomarker Genes Using Hierarchical Clustering

Abstract

Assessment of drugs toxicity and associated biomarker genes is one of the most important tasks in the pre-clinical phase of drug development pipeline as well as in the toxicogenomic studies. There are few statistical methods for the assessment of doses of drugs (DDs) toxicity and their associated biomarker genes. However, these methods consume more time for computation of the model parameters using the EM (Expectation-Maximization) based iterative approaches. To overcome this problem, in this paper, an attempt is made to propose an alternative approach based on hierarchical clustering (HC) for the same purpose. There are several types of HC approaches whose performance depends on different similarity/distance measures. Therefore, we explored suitable combinations of distance measures and HC methods based on Japanese Toxicogenomics Project (TGP) datasets for better clustering/co-clustering between DDs and genes as well as to detect toxic DDs and their associated biomarker genes. We observed that Word’s HC method with each of Euclidean, Manhattan and Minkowski distance measures produces better clustering/co-clustering results. For an example, in case of glutathione metabolism pathway (GMP) dataset LOC100359539/Rrm2, Gpx6, RGD1562107, Gstm4, Gstm3, G6pd, Gsta5, Gclc, Mgst2, Gsr, Gpx2, Gclm, Gstp1, LOC100912604/Srm, Gstm4, Odc1, Gsr, Gss are the biomarker genes and Acetaminophen_Middle, Acetaminophen_High, Methapyrilene_High, Nitrofurazone_High, Nitrofurazone_Middle, Isoniazid_Middle, Isoniazid_High are their regulatory (associated) DDs explored by our proposed co-clustering algorithm based on the distance and HC method combination Euclidean: Word. Similarly, for the PPAR signaling pathway (PPAR-SP) dataset Cpt1a, Cyp8b1, Cyp4a3, Ehhadh, Plin5, Plin2, Fabp3, Me1, Fabp5, LOC100910385, Cpt2, Acaa1a, Cyp4a1, LOC100365047, Cpt1a, LOC100365047, Angptl4, Aqp7, Cpt1c, Cpt1b, Me1 are the biomarker genes and Aspirin_Low, Aspirin_Middle, Aspirin_High, Benzbromarone_Middle, Benzbromarone_High, Clofibrate_Middle, Clofibrate_High, WY14643_Low, WY14643_High, WY14643_Middle, Gemfibrozil_Middle, Gemfibrozil_High are their regulatory DDs. These results are validated by the available literature and functional annotation.

Robust Hierarchical Co-clustering to Explore Toxicogenomic Biomarkers and Their Regulatory Doses of Chemical Compounds

“…It is mentioned earlier that the combinations euclidean: ward, manhattan: ward and minkowski: ward of distance and HC methods are suitable for clustering/co-clustering toxicogenomic data (Hasan et al, 2019b). In this paper, we have also claimed that these combinations of distance and clustering methods are very sensitive to outlying observations in the dataset.…”

“…Many authors' consensus is that identification of subsets of DCCs which have similar mechanism of action over the respective subsets of genes as well as identification of the biomarker genes and their regulatory DCCs are the main objectives of toxicogenomic studies as well as drug development process (Madeira and Oliveira, 2004;Afshari et al, 2011;Nyström-Persson et al, 2017;Hasan et al, 2018;Hasan et al, 2019b). The HC algorithm is more popular and widely used method which can be used for clustering genes and DCCs.…”

“…Furthermore, the LPHVM creates artificiality considering absolute value of the dataset. In this aspect Hasan et al, 2019b proposed a co-clustering algorithm and showed that the algorithm co-cluster the genes and DCCs efficiently to explore toxicogenomic biomarkers and their regulatory DCCs. The author also proved that the combinations of distance HC methods (euclidean: ward, manhattan: ward and minkowski: ward are suitable combinations for clustering genes and DCCs of toxicogenomic data.…”

“…Observing all the plots given in Figure 2 and in the supplementary material we can conclude that the distance and HC method combination manhattan:ward.D produces lower ER, as well as more stable results, compare to the others proposed approaches (euclidean:ward.D, euclidean:ward.D2, manhattan:ward.D2, minkowski:ward.D and minkowski:ward.D2) for clustering genes or DCCs is considered as the best approach for toxicogenomic data. Co-clustering (clustering genes and DCCs simultaneously) genes and DCCs is one of the important objectives of toxicogenomic studies as well as drug development processes (Afshari et al, 2011;Hasan et al, 2018;Hasan et al, 2019b). Therefore, we have proposed the RHCOC algorithm for coclustering genes and DCCs to explore toxicogenomic biomarkers and their regulatory DCCs.…”

“…Nevertheless, all of them are not equally suitable for clustering genes and DCCs of toxicogenomic data. Recently, a method has developed that the combinations of distance and clustering methods euclidean: ward, manhattan: ward and minkowski: ward are suitable for clustering as well as co-clustering (clustering row (gene) and column (DCCs) entities of a data matrix simultaneously) genes and DCCs (Hasan et al, 2019b). However, the algorithm has two important shortfalls.…”