I agree with Lila - you should not perform differential expression analysis on FPKM counts. If you don't believe me, then take it from the developer of limma, where some suggestions are also made: https://support.bioconductor.org/p/56275/#56299
The Total Count and RPKM [FPKM] normalization methods, both of which are still widely in use, are ineffective and should be definitively abandoned in the context of differential analysis.
Also, by Harold Pimental: What the FPKM? A review of RNA-Seq expression units
The first thing one should remember is that without between sample normalization (a topic for a later post), NONE of these units are comparable across experiments. This is a result of RNA-Seq being a relative measurement, not an absolute one.
Remarkably, I still see publications coming out where people are comparing groups of samples based on FPKM counts, even though this makes no sense. As an example, FPKM of 10 in one sample may be the equivalent of 50 in another, due to the way that FPKM counts are produced, i.e., with no cross-library / sample normalisation.
Yes you can use limma although It is not as good as if you had counts - so double check you cannot get the counts and try writing the people behind the data to see if they will provide the counts (most people are quite friendly in my experince). If you cannot get the counts you can log2 transform the FPKM values (use a pseudocount of 1) and use the limma trend approach described on page 71 of the limma vignette. For a comparison of limma-trend vs limma-voom take a look at this article.
Hope this helps