I am using DESeq2 to process the bacterial (metatranscriptome) response between 2 conditions. Each condition includes 5 biological replicates. After finishing the DESeq2 analysis I generated a volcano plot and saw that almost all the genes with the lowest padj actually have very low log2foldChange values. All replicates in the DPS condition have zero counts for those genes while DVM replicates mostly show quite high values. I would think that comparing between zeros in the DPS condition should produce a higher log2fold change in the DVM condition. Could you please share your opinion why would this happen?