Hi Subramaniyam, 

It would better if you use normalized read count rather than using raw read count.This is because on normalizing your read counts you may got some what different results as compared to one calculated using raw read count. This type of inferences are seen because the normalization factor i.e calculated to normalize your raw read count int RPKM or FPKM or TPM depends on both the transcript specific read count and total read count. You can use the formula for calculating RPKM as follows:

C = Number of reads mapped to a gene

N = Total mapped reads in the experiment

L = exon length in base-pairs for a gene

RPKM = (10^9 * C)/(N * L)

There might be some deviation in values of log2 fold change calculated manually by using relative RPKM values and the one calculated using DE tools. You can also use statistical edgeR package to perform differential gene expression analysis using the read count matrix created by Htseq tool.

hi, you should normalize your counts by some normalization methods for example quantile normalization method by package lima or min-max normalization method depending on your job

You should use normalized reads count measures such as RPKM or TPM. 

Hi,

No, raw reads counts are not going to give you reliable estimation of differential expression. First because samples will have different library sizes. But RNA composition also need to be adjusted for, etc.The differential analysis is not trivial, but packages and software are there to help.

I am not familiar with Gfold, but check the user guide to check what data you need to provide. edgeR for example requires raw counts, not RPKM or FPKM.

Good luck,