Which counts are to be considered when using Noiseq?

Question

[Hyper_Odin]

I am performing normalization using Noiseq, but I am not sure which counts to use. I performed normalization using HTSeq counts and from Kallisto counts and the results seem to be different. When I compared the results, some of the genes seem to be fine but for most of the genes, fold change and probability are quite different. I am performing TMM normalization because we don't have any replicates.