Hey team, Had a question about the custom panel training. I currently have matched tumor/normal WES as well as tumor bulk RNA seq (melanoma). If my main goal is looking at results from LILAC to see if there is any sort of loss of heterozygosity as well as any effects to the antigen presentation pathway does it make sense for me to do this training? I was able to run nextflow in targeted mode with the tso500 panel. However, when looking at the documentation of the README_TARGETED it seems a bit confusing when trying to use a custom bed / reference. Was mainly curious to see from what you guys see, if there is a drastic difference in results when looking at WES custom panel vs tso500. Also in terms of the training set, I'm assuming for the copy number it needs to be a pool of tumors? For SAGE does it need to be a pool of normals so 2 different training sets? If I have about 100 tumor samples is there any recommendation on how many I should use to train (does it make sense to use all 100 or just a subset? Thanks!

Hi @Kairi Tanaka, you’ll need to generate/train targeted resource files for your specific WES panel since each panel will have different biases that need to be controlled Analysing data from one panel with resource files from another panel (e.g. WES data with TSO500 panel resource files) will likely give you low quality results The instructions for training panel resource files has been recently updated, and the upcoming oncoanalyser 2.2.0 release will include functionality to perform most of this for you as well The documentation recommends using 20 or more representative samples for training, so I would say selecting 20 samples would be a good starting point