When I ran PureCN using matching tumor/normal pairs, most of my samples were flagged as having a noisy log-ratio. However, all samples successfully ran through PureCN and none were flagged as Failed.
I then ran PureCN with --normal removed, so it would create a representative normal sample itself. You had indicated this tended to produce better results. Now, I've had several samples fail with a fatal error saying that no solution could be found. These samples all seem to be at the lower end of our coverage.
Below is a screen shot of a plot showing mean coverage of each sample. The samples that initially failed were the JES-61 and IBG-2-CG-178 samples. I removed those from the analysis and tried again, and then sample PC-60T3 failed, so I removed it. Now, sample CH-59T1 has failed. Just what causes this failure, when it worked with matched normals? Is this solely a problem with low coverage, or could there be another problem here?