search

Nik-Zainal-Group / signature.tools.lib

R package containing useful functions for mutational signature analysis
Other
80 stars 26 forks source link

Rearrangement signature R5 in pancreas #35

Closed felixbeaudry closed 2 years ago

felixbeaudry commented 2 years ago

Hi there! Thanks for a very cool package! I have a quite specific question: I am running the rearrangement signatures SignatureFit_withBootstrap_Analysis for a pancreas sample getOrganSignatures(organ="Pancreas",typemut = "rearr") and then converting to reference convertExposuresFromOrganToRefSigs and I'm coming across a surprising result: my sample has a lot of Pancreas_C, which in Fig 3 of Degaspari et al 2020 (https://www.nature.com/articles/s43018-020-0027-5) is colored as R5 but in the conversion matrix (signature.tools.lib/data/2019_01_10_ConversionMatrix_rearr.tsv ) Pancreas_C is weighted as R9 (weight=1). Since RefSig R5 is associated with BRCAness but not R9, I am getting much less confidence on BRCAness for my pancreas samples when running HRDetect_pipeline. May I ask how the conversion matrix was made compared to the manuscript figure?
Thanks for your help! Felix

andreadega commented 2 years ago

Hi thanks for the compliment!

Sorry about the confusion of Figure 3. The colouring is based on the cosine similarity to the six breast cancer rearrangement signatures that were reported in 2016. The legend of Figure 3 says: "...colored according closest rearrangement signature derived in Nik-Zainal et al.[12], based on cosine similarity.".

In fact, in Pancreas we do not find R5, but only R9, and we consider them somewhat equivalent, because they are both short deletion signatures (see in Figure 6). In fact, for HRDetect we consider both R5 and R9 exposures as the HRDetect input SV5. If you have run HRDetect_pipeline, then the SV5 exposures should include the R9 exposures, and you should be able to see the R9 exposures in the data_matrix returned by the HRDetect_pipeline.

andreadega commented 2 years ago

Closed for inactivity.