Hello Romit, I don't remember exactly how edgeR works with a different number of duplicates for several samples. If you use DESeq2, which is a more widely used derivative of edgeR, the you can for sure use you data as they are structured without any problem. Just keep in mind that the biological dispersion will be a bit less well evaluated with duplicates than with triplicates, so that you may want to choose a higher p-value / FDR threshold when you select DE genes with your sample in duplicates being one of the condition. Good luck!

Thanks Fabrice, further, I want to know whether using such approach is acceptable by journals. 

As long as you can explain your experimental scheme and the choices you made, I think your data are acceptable as such. Of course, from the statistician point of view, it's always better to compare many samples each with a lot of replicates, but most reviewers are quite aware of the time and money costs of such experimental plannings!