Figure 2

Visualization of noise reduction performance with a simulated data set. In each panel, the thin rectangular image on the left shows a complete view of the 2000 rows × 60 columns matrix, and the large square image on the right shows a close-up image of the red-enclosed part of the matrix. (A) Simulated signal matrix. (B) Simulated data matrix, in which random noise was added to (A). (B) was subjected to RSPR-NR (C), PCA with PCs1-3 kept (D), PCA with PCs1-9 kept (E), or PCA with PCs1-26 kept (F). See Figure 3 for the reasons that these numbers of top PCs were kept in PCA. Yellow, gray, and blue show positive, zero, and negative values, respectively. The parameters used for the simulated data set were 30 large and 60 small signal blocks and a noise variance ratio of 0.01. The parameters used in RSPR-NR were a subset row number of 200, a repeat number of 20, and an FDR of 0.0316.