Hi, there! 

I used a transposed otu table, and sample data from phyloseq to construct a DESeq object. 

However, in my otu table, there are zeros in almost each column, which prevent DESeq to do the log transform. I noticed that if I just simply add 1 to all values, I can bypass the issue. But there are several discussions indicating we shouldn't add 1 to the otu table. 

I am just wondering, what will the best way to fix it then?

Thank you very much! 

Zeros present for every OTU are only a problem for the default size factor estimation. You can use estimateSizeFactors() with the argument type="poscounts" in the latest version of DESeq2. This method has evolved from phyloseq forums, and is also tested in this preprint

http://www.biorxiv.org/content/early/2017/06/30/157982