search

aeeckhou / shallowHRD

This method uses shallow Whole Genome Sequencing (sWGS) and the segmentation of a genomic profile to assess the Homologous Recombination Deficiency of a tumor based on the number of Large-scale Genomic Alterations (LGAs).
30 stars 13 forks source link

cMAD question #19

Closed carolinehey closed 6 months ago

carolinehey commented 6 months ago

I have a question about the cMAD that i hope you can anwser. I the shallowHRD_hg19_1.13_QDNAseq_no_chrX.R code i found this snippet: image However, further up in the script i see thresholds matching what is specified in the Bioinformatic report: image

Are these cMAD/QC_MAD_point_corrected supposed to be the same, and classified as bad >0.50 or 0.05 ? Finally, there is also a threshold cMAD<0.14 is good.

I am a bit confused and hope you can clarify this.

Kind regards, Caroline

aeeckhou commented 6 months ago

Hello Caroline,

I have included in the very beginning the cMAD feature to use it as a mesure of quality. That part of the code is from very old time. The cMAD variable is deprecated (first snipplet) to the _QC_MAD_pointcorrected calculation. With it, I defined the first rule as previously described in the Bioinformatics report.

It is the variable _QC_MAD_pointcorrected that is used in the final table, not the cMAD !

a = c("CNA cut-off value", "Number simulations for cut-off", "Number successfull simulations", "MAX2", "cMAD",  "Number LGAs 10Mb", "HRD")
b = c(THR, format(nb_simulations,scientific = F), number_positive, round(MAX2,3), QC_MAD_point_corrected,   number_LGAs, HRD)

This criteria was somewhat satisfying at the time but I changed since then the pipeline, relying on random sampling to insure a better CNA cut-off detection.

For this version 1.13 (not the 2.0), good quality indicators are :

  1. The number of succesfull simulations devided by the total number of simulations. < 10 % was associated with bad qualities.
  2. The MAX2 value is a good surrogate marker of tumor content. Above 0.16, the tumor content is probably good.

I would rely more on them. hope I clarified the "cMAD inclusion part" that should not be there.

Is it clear for you ?

Best regards, Alexandre Eeckhoutte

carolinehey commented 6 months ago

Thank you very much for your answer, it is much more clear to me now :)