Skip to main content
Fig. 6 | BMC Bioinformatics

Fig. 6

From: An individualized predictor of health and disease using paired reference and target samples

Fig. 6

Biomarkers selected by reference-aided predictor. The top figures in (a) show the genes selected by the proposed reference-aided predictor with selection frequency ≥70 % for the 4 different score functions for states 2,3,4,5. The value of the classifier weights for each of the score functions are shown as yellow bars (weights applied to reference sample R) and green bars (weights applied to target sample T). Note that genes having yellow and green bars of opposite sign are contrasting information in R vs T while genes having these bars with the same sign are reinforcing information in R and T. The bottom figures in (b) show the expression of the genes shown in (a) averaged over the uninfected subjects (left) and infected subjects (right). The expression levels are normalized such that the maximum and minimum of each gene achieve 1 and −1 respectively. Let the averaged expression at time t be z(t), the maximum of z(t) be z max , and the minimum be z max . The normalized expression levels are computed as \(\tilde z(t)= 2\times \left [z(t)-0.5\times (z_{\textit {max}}+z_{\textit {min}})\right ]/(z_{\textit {max}}-z_{\textit {min}})\)

Back to article page