I have 17k peptides and 96 samples imported from dia-nn report. No preprocessing except of filterNonProteotypicPeptides and filterCompoundProteins was done.

dpc() gives 0.63 and dpcCN() gives 1.05. Which of them should I use in limpa::dpcQuant()?

I expect that very few samples may be outliers.

I tend to use dpcCN(), because it more robust than dpc() to experiments with lots of samples and big differences between samples. You could also try dpcON(, robust=TRUE), which is a newer, faster and more robust version of dpc(). We are in the process of phasing out the original dpc() in favour of dpcON().

We find that limpa is pretty robust to the exact dpc.slope that is used, and you will probably get pretty consistent DE lists using either 0.63 or 1.05.

Please see the example limpa analyses at https://github.com/smythlab/limpa/. The renal cancer case study provides an example of filtering and normalization.