Skip to main content
Figure 3 | BMC Cancer

Figure 3

From: Statistical techniques to construct assays for identifying likely responders to a treatment under evaluation from cell line genomic data

Figure 3

Out-of-bag error computation schematic. A schematic to illustrate the out-of-bag error computation procedure to determine the optimal regularization parameter value(s) for the Lasso, RDA, or SNSS, in the context of the dacetuzumab sensitivity classification case study. We repeat the following B times. We sample the N cell lines in the full cell line data N times, thus creating a bootstrap data set that will serve as our training sample. The approximately 37% of the cell lines left out of the bootstrap data set as a result of the sampling procedure will serve as an out-of-bag sample. We perform feature selection and construct a classifier using the bootstrap data, which we use to classify each of the cell lines in the out-of-bag sample as dacetuzumab sensitive (0), semi-sensitive (1), or resistant (2). For each cell line, we determine its final out-of-bag sensitivity classification by tallying up the classifications we obtained during the times the cell line appeared in the out-of-bag sample and selecting the sensitivity according to a majority vote. We compare the final out-of-bag classifications to the true sensitivities to determine the out-of-bag error.

Back to article page