Skip to main content

Advertisement

Figure 6 | Genome Medicine

Figure 6

From: Computational purification of individual tumor gene expression profiles leads to significant improvements in prognostic prediction

Figure 6

Comparison of ISOpure-predicted and pathologist reported percentage cancerous tissue. (A) Scatter plot of ISOpure predictions against the average pathologist estimates on a subset of 20 lung tumors and three blinded healthy lung tissues from the Bhattacharjee dataset. The size and color of each point indicate the difference between the pathologists' estimates. The blue region indicates where the ISOpure predictions were within 13.7% of the average pathologist estimate. (B) Median-centered expression levels of a random selection of 100 genes in 50 patients of the Bhattacharjee dataset, before ISOpure purification. (C) Same genes and patients as in (B), but expression levels were from the ISOpure cancer profiles. (D) Scatter plot of ISOpure predictions against a single pathologist on the Wang dataset of 109 prostate tumor samples. The black dashed line indicates the linear regression model that minimizes the sum of squared errors. (E) Correlation of ISOpure estimates and the average of the two pathologists' estimates on the same 23 samples as in (A), depicted as a function of the number of normal samples made available to ISOpure. Each point represents a random selection of normal samples of the given size. (F) Same as in (E), but correlation was measured for the 109 samples in the Wang dataset.

Back to article page