Let's say I have an RNA seq experiment where I compare three conditions: untreated (UT), infected (HT), and mock-infected (MT) for both infants and adults. I ask the following questions:

1. What are the differentially expressed genes between UT infants and UT adults, etc.?

   contrast <- eval(c("condition",  "infant_HT", "adult_HT"))  # same syntax for the other conditions
   sample_data <- sample_data %>% drop_na()
   se <- tximeta::tximeta(sample_data, txOut=T, type="salmon")
   # summarize object to gene level
   gse <- summarizeToGene(se)
   dds <- DESeqDataSet(se = gse, design = ~condition)
   # Only keep genes that have 6 or more reads across all the samples
   keep <- rowSums(counts(dds)) >= 6
   dds <- dds[keep,]
   dds <- DESeq(dds, test = "Wald")
   res <- results(dds, contrast=contrast)
   

2. What genes are differentially expressed between all 3 groups independent of age?

   sample_data <- sample_data %>% drop_na()
   se <- tximeta::tximeta(sample_data, txOut=T, type="salmon")
   # summarize object to gene level
   gse <- summarizeToGene(se)
   dds <- DESeqDataSet(se = gse, design = ~condition)
   # Only keep genes that have 6 or more reads across all the samples
   keep <- rowSums(counts(dds)) >= 6
   dds <- dds[keep,]
   dds <- DESeq(dds, test = "LRT", reduced ~ 1)
   res <- results(dds,)
   

   In question number 1 I determine DEGs from UT, HT, and MT because I am interested in seeing how these differ with age (infant and adult). Additionally, I am also curious about what genes are unique to UT, HT, MT and which are shared across the three treatment conditions. At this point, I have 3 DEG tables with respect to age and 4 lists of genes (composed of unique or overlapping genes). In question 2 I am curious about what these genes are and to what foldchange and significance they differ. To accomplish this, I have to reevaluate DEGs (see code block under question 2); however, when I do this, I find that the overlapping genes between the 3 conditions I calculate independently are not equal to the DEGs of the reevaluated data.

This makes sense, but how can I determine the metrics for the overlapping genes if reevaluating DEGs leads to a different result? Am I not able to assess unique DEGs across different treatment conditions?

Unfortunately, I have to reserve my time on the support site for software related questions. For consultation about statistical design and interpretation of results, I recommend working with a local statistician or someone familiar with linear models in R. Note that, in general not all combinatorial hypotheses can be expressed as a single contrast and so it makes sense to just combine across multiple results tables from different contrasts.