Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Drug candidate identification based on gene expression of treated cells using tensor decomposition-based unsupervised feature extraction for large-scale data

Fig. 2

An overview of the analysis using TD-based unsupervised FE. Top left: Gene expression tensor xijℓ of dose dependence mode (i), compound mode (j), and gene mode (ℓ). Top right: Using the TD, xijℓ, was decomposed to the tensor product of core tensor \(G_{k_{1},k_{2},k_{3}}\), dose dependence matrix \(x_{k_{1},i}\), compound matrix \(x_{k_{2},j}\), and gene matrix \(x_{k_{3},\ell }\). Bottom right: Because the second component of dose dependence mode shows linear dose dependence (Additional file 2), and cumulative contribution of the core matrix up to the sixth components exceeds 95% of the total contribution, core matrix \(G_{k_{1}=2,k_{2} \le 6,k_{3} \le 6}\) is considered for FE. Bottom left: Outlier compounds (they correspond to ‘inferred compounds’ in Table 1) and outlier genes (they correspond to ‘inferred genes’ in Table 1) are identified within the space restricted with \(x_{k_{2} \le 6,j}\) and \(x_{k_{3} \le 6,\ell }\), respectively

Back to article page