Skip to main content
. 2022 Mar 8;12:4133. doi: 10.1038/s41598-022-07685-4

Figure 4.

Figure 4

Semi-supervised learning for the tissue section imputation. (a) Box plots show Pearson’s correlation coefficients between the measured and predicted gene-expression levels of 21 breast cancer-related microenvironment markers. Left box plot displays the results of semi-supervised learning, which showed increasing Pearson’s correlation coefficients. Middle and right box plots show the semi-supervised learning results with permutated and randomized values. For the box plot, the box indicates the first and third quartiles; horizontal center line marks the medians; upper whisker extends from the hinge to the highest value that is within 1.5× interquartile range (IQR) of the hinge; lower whisker extends from the hinge to the lowest value within 1.5× IQR of the hinge; and data were plotted as points. Black lines between boxes connect the same gene. (b) Box plots show Pearson’s correlation coefficients between the measured and predicted gene-expression levels of 21 breast cancer-related microenvironment markers. Three types of image sets were compared for semi-supervised learning, namely sections A–C (red); data from The Cancer Genome Atlas (TCGA) (blue); and ImageNet data (green). For the box plot, the box indicates the first and third quartiles; horizontal center line marks the medians; upper whisker extends from the hinge to the highest value that is within 1.5× IQR of the hinge; lower whisker extends from the hinge to the lowest value within 1.5× IQR of the hinge; and data were plotted as points.