I'm running DESeq2 on a small RNA seq experiment. The coldata has 87 samples total and begins like this:

My two main comparisons, Lean vs Obese and PCOS vs non-PCOS work fine, which I can summarize with this:

dds <- DESeqDataSetFromMatrix(countData = cts_df_ordered,
  colData = coldata_ordered_trim,
  design = ~ Particle + BMI + Diagnosis)
dds$Diagnosis <- factor(dds$Diagnosis, levels = c("PCOS","non-PCOS"))
dds$BMI <- factor(dds$BMI, levels = c("Lean","Obese"))
dds$Particle <- factor(dds$Particle, levels = c("MP","FFD","EV"))
dds <- DESeq(dds)

resDIAG <- results(dds)
resresOrdered <- resDIAG[order(resDIAG$pvalue),]
summary(resDIAG)
sum(resDIAG$padj < 0.05, na.rm=TRUE)

resBMI <- results(dds, contrast=c("BMI","Lean","Obese"))
summary(resBMI)
sum(resBMI$padj < 0.05, na.rm=TRUE)

Then I tried to subset the dds to make these same two comparisons but within each particle, starting with MP:

dds_MP <- dds[, dds$Particle %in% c("MP")]
dds_MP$Particle <- droplevels(dds_MP$Particle)
dds_MP <- DESeq(dds_MP)

Giving the error:

using pre-existing size factors
estimating dispersions
found already estimated dispersions, replacing these
Error in `contrasts<-`(`*tmp*`, value = contr.funs[1 + isOF[nn]]) : 
  contrasts can be applied only to factors with 2 or more levels

I'm also looking to make PCAs which only include the particle type of interest, which I was doing for the complete dataset using the below code:

vsd <- vst(dds, blind=FALSE)
plotPCA(vsd, intgroup=c("BMI","Diagnosis"))

How can I subset the 3 particle types (MP, FFD, EV) and compare PCOS v non-PCOS and Lean vs Obese within particle type followed by PCA? Thanks in advance for your time and help.

From quickly scanning your code, it looks like you need to change the design(dds) before you re-run DESeq() on the subset. You still have a factor in there even though that factor only has one level, and so it doesn’t make sense for building coefficients. DESeq2 does some things (mostly checks) on the user’s behalf, but it doesn’t change the design(dds) for you, as this is a important user-defined part of the analysis.