Closed JuseTiZ closed 1 month ago
Hi JuseTiZ,
Thank you for your question.
The discrepancy you observed between cosmic_fit
and denovo_fit
is indeed due to differences in the processing methods and parameter settings of the two functions.
cosmic_fit
: This function is designed to be more stringent and only assigns high-confidence signatures to the samples. It aims to minimize overfitting by applying stricter criteria for signature assignment.denovo_fit
: In contrast, denovo_fit
operates with less stringent cutoffs, allowing it to identify and assign signatures that may be present at lower confidence levels. This approach can lead to the detection of additional signatures, like SBS19 in your case, that cosmic_fit
might not have identified.Usually, if you have fewer samples and wish to fit with the existing databases of reference signatures, like COSMIC, cosmic_fit
is better. If you are fitting with de novo signatures, which could be stemming from noisy data or sensitive to parameters of signature extraction, you can use denovo_fit
.
I hope this helps clarify the differences and how to approach your analysis. Feel free to reach out if you have any more questions via email!
Best, Ravi
When using
cosmic_fit
to assign mutational signatures to my samples, I identified SBS1, SBS5, and SBS19.I then extracted these three signatures to refit using
denovo_fit
. However, I noticed differences in the distribution of these signatures. For example, SBS19 appeared in samples where it had not been detected previously withcosmic_fit
.Is this discrepancy due to differences in the processing methods of the two functions? If I want to compare the changes in fit after adding some signatures, should I compare the results of two
denovo_fit
runs (before and after adding the signatures), or should I compare the results ofdenovo_fit
(after adding the signatures) withcosmic_fit
?Looking forward to your reply!