Multiple tests between multiple conditions and covariates with sleuth

Question

[danjs116]

I'm attempting to run a differential expression analysis on RNA-seq data from cells in 4 different conditions:

• Tissue 1 (T1) infected with Virus 1 (V1)
• Tissue 1 (T1) infected with Virus 2 (V2)
• Tissue 2 (T2) infected with Virus 1 (V1)
• Tissue 2 (T2) infected with Virus 2 (V2)

I've constructed a mapping table annotating abundance files (derived from a kallisto alignment), to my samples, with information about the tissue and virus type of the sample. I've set the reference level of tissue to be T1, and virus to be V1. After this, I've used the following script to produce 5 different comparison tests:

mart <- useMart(biomart = "ENSEMBL_MART_ENSEMBL",
                       dataset = "hsapiens_gene_ensembl",
                       host = "www.ensembl.org")
t2g <- getBM(attributes = c("ensembl_transcript_id", "ensembl_gene_id",
                                   "external_gene_name"), mart = mart)
colnames(t2g) <- c("target_id", "ens_gene", "ext_gene")
so <- sleuth_prep(metadata, target_mapping = t2g, aggregation_column = "ens_gene")

so <- sleuth_fit(so, ~virus + tissue + virus:tissue, 'test1')
so <- sleuth_fit(so, ~virus + tissue, 'test2')
so <- sleuth_fit(so, ~virus, 'test3')
so <- sleuth_fit(so, ~tissue, 'test4')
so <- sleuth_fit(so, ~1, 'test5')

so <- sleuth_lrt(so, 'test5', 'test3')
so <- sleuth_lrt(so, 'test4', 'test1')
so <- sleuth_lrt(so, 'test5', 'test4')
so <- sleuth_lrt(so, 'test3', 'test1')
so <- sleuth_lrt(so, 'test2', 'test1')

#Effect of virus in the T1
sleuth_table_1 <- sleuth_results(so, 'test5:test3', 'lrt', show_all = FALSE)

#Effect of virus in the T2
sleuth_table_2 <- sleuth_results(so, 'test4:test1', 'lrt', show_all = FALSE)

#Difference between T2 and T1 with V1
sleuth_table_3 <- sleuth_results(so, 'test5:test4', 'lrt', show_all = FALSE)

#Difference between T2 and T1 with V2
sleuth_table_4 <- sleuth_results(so, 'test3:test1', 'lrt', show_all = FALSE)

#Difference between T2 and T1 response to V2
sleuth_table_5 <- sleuth_results(so, 'test2:test1', 'lrt', show_all = FALSE)

In all cases, the number of differentially expressed genes is much lower than expected, and lower than the number of DEGs identified using Deseq2 to perform these tests (after gene-level assignment by tximport). Any help with this would be much appreciated.