Leveraging TCGA gene expression data to build predictive models for cancer drug response

BMC Bioinformatics

Table 2 Number of clusters and mean accuracy for pan-cancer models

	Fluorouracil			Gemcitabine
Clustering Method	# Considered Clusters^b	# Selected Clusters^c	Average Accuracy	# Considered Clusters^b	# Selected Clusters^c	Average Accuracy
Optcluster (Clara) with RF^a	204	32	84.1%	192	50	82.3%
Optcluster (Clara) with SVM^a	204	32	81.0%	192	50	71.5%
Optcluster (Clara) with logistic regression	204	32	77.0%	192	50	73.0%
Optcluster (Clara) with RF and demographics	204	32	83.6%	192	50	79.1%
Model Validation (Clara)	204	32	52.9%	192	50	82.1%
Model Validation (Clara with Tuning)	204	32	52.9%	192	50	85.7%

^aRF is for random forest; SVM is for support vector machine
^bNumber of considered clusters represents the number of clusters entered into random forest for variable importance ranking
^cNumber of selected clusters is number of clusters selected by random forest for classification

ISSN: 1471-2105