Reproducing images from Cristiano et al. 2019

[agalitsyna]

Hi! Our team tries to reproduce Figures 3b-c from Cristiano et al. 2019 paper. This is the repo mentioned in the paper, so I guess it's the right place to discuss concerns about the reproducibility of the images from the paper.

First, I try to reproduce Fig. 3b based on the output of delfi_scripts. I plotted ratio.corrected2 from bins_5mbcompartments.csv but had to subtract the mean from each sample to normalize the tracks to zero. Is this the right normalization used for that plot, or should I use something more sophisticated? I believe it is not clarified in the paper, and the code for the generation of Fig. 3b is not in delfi_scripts.

Next, I try to reproduce Fig. 3c (guessing that the subtraction of mean is enough for normalization). I have not found the code to generate this image either. I obtain the windows with more than 10% of the cancer samples with a fragment ratio of more than three standard deviations from the median healthy fragment ratio. I obtain the results that deviate substantially from Fig. 3c. Some samples have radically more significantly differing bins; some have only a few.

I wonder what might be the reason for this discrepancy. For example, how the deviation from the median healthy is calculated? Is it based on the whole-genome deviation or bin-wise? Is it calculated for each cancer type independently, or for all the samples?

I will appreciate it if you can add the scripts to generate the Figures 3b-c from the paper or guide towards the appropriate reproducibility of the results of your research.