I am trying to compare treatment (A) vs my control (C). A heavily downregulates genes, but for some reason my volcano plot shows that it is upregulated when compared to the control. The samples were not swapped (I've checked) and I got the correct distribution (downregulation) once, but since then have gotten the reverse. What is going wrong?

dds <- DESeqDataSetFromMatrix(countData = my_cts,
                          colData = my_coldata,
                          design = ~ condition)
dds

#prefilter row sums with less than 10 reads
keep <- rowSums(counts(dds)) >= 10
dds <- dds[keep,]

#factor levels
my_coldata$condition <- factor(my_coldata$condition, levels = c("C","A"))

#do i need these?
    #dds <- estimateDispersions(dds)
    #dds <- nbinomWaldTest(dds) 

#run DESeq2
dds <- DESeq(dds)

#check coefficients estimated by DEseq
resultsNames(dds)

#obtain results
res <- results(dds, contrast=c("condition","A","C")) #order is treatment, numerator, denominator

#shrink LFC
resLFC <- lfcShrink(dds, coef = 2, type = "apeglm") #coef is set to "condition A vs C"

#make volcano plot (the result that looks flipped) 
par(mfrow=c(1,1))
# Make a basic volcano plot
with(res, plot(log2FoldChange, -log10(pvalue), pch=20, main="Volcano plot", xlim=c(-4,4)))
with(subset(res, padj<.01 ), points(log2FoldChange, -log10(pvalue), pch=20, col="blue"))

Any advice is much appreciated.

A very similar question was asked recently on the support site.

When you pass an object to a constructor function, it makes a copy. So later when you modify it, it doesn't affect the dds.

Here

#factor levels
my_coldata$condition <- factor(my_coldata$condition, levels = c("C","A"))

You've modified my_coldata but since dds has a copy, you haven't modified dds.