hi, i am grad student who is mid of master's degree. one of my project is revealing some genes that differentially expressed between cancer that has small size and do not lymph node metastasis and cancer that has small size and do lymph node metastasis. i'd done sampling from 10 of each ffpe blocks, and done RNA-seq, pre-processing. and i've just ran DESeq2 for find DEG and i got only 4 significantly differently expressed genes from results (using fdr alpha = 0.1 which is default in DESeq2) but 2 of that 4 DEG have high intra-condition variance and i throw away that high variance genes. so there are only 2 DEG.
here is my question, the condition between two groups, is highly similar. so i inferred there are small DEG in all ~50000(coding + non-coding) genes. so like, normal vs cancer DEG analysis shown lots of DEG before p adjusted. so many of them survived after p-value adjustment. but in this case, there are small DEG before adjustment cus of similar condition. so after FDR all DEG(before adj.) disappear or really small subset of genes survival(most of them, seem false positive, have high variance or too much LFC like log2 fold change = 20)
in this case, is there any best practice for DEG analysis? i'm doing wilcoxon rank sum test and cut genes with hard criteria. like, p-value < 0.001, intra-sample sd <0.5, LFC >=1. as a results i take approx. 35 genes of differentially expressed that have big LFC, and small intra condition SD. but, i think it is not the best methods.
please help. and many many thanks in advance.
-han sai-
