6  Differential Expression Analysis II

Combined analysis with batch as a covariate

Published

October 29, 2025

6.1 Handling Batch-Treatment Confounding

When treatment is nested within batch, any observed difference between Cre-AAV and GFP could potentially be attributed to either true biological effects of the treatment or technical batch effects. However, unlike complete confounding, our design does allow for batch adjustment because GFP control samples exist in both batches.

A model with ~ batch + Sex + group is mathematically valid: the batch effect can be estimated from GFP samples present in both batches, and the treatment effect is then estimated as the Cre-AAV versus GFP difference within Batch 2 after accounting for the batch effect.

6.2 Pre-filtering low count genes

Genes with very low read counts were excluded to minimize technical noise. After filtering for a minimum of 20 counts in at least five samples, approximately 19011 genes were retained for downstream analysis.

6.3 Differential expression analysis between Cre-AAV and GFP control groups in switch mice

Differential expression analysis was performed using DESeq2 with a design formula that explicitly modeled batch effects: ~ batch + Sex + group. GFP Control was set as the reference level for treatment comparison.

Our design involves the same mice measured across batches (paired structure) nested within a batch-treatment confound. The ideal model would be a mixed-effects model with random effects: Treatment + (1|MouseID) + (1|Batch) DESeq2’s standard framework assumes all samples are independent and cannot model the correlation between repeated measurements of the same biological individual through random effects (e.g., mouse-specific baseline expression). Retaining both measurements would violate this independence assumption. While removing Batch 1 duplicates would resolve this, we chose to collapse cross-batch measurements by averaging to preserve all sequencing data. This approach treats each mouse as a single biological unit, though it reduces variance by averaging over real batch-to-batch measurement variability. While this reduces variance by averaging over real batch-to-batch variability (potentially anti-conservative), PCA showed minimal batch effects.

Significance thresholds were set at FDR < 0.05 and |log₂FC| > 1, with subsequent filtering to remove pseudogenes, unannotated loci, mitochondrial genes, and ribosomal protein genes.

6.3.1 Liver

In liver tissue from all mice (males and females combined), we identified 121 significantly differentially expressed genes between Cre-AAV and GFP control animals (FDR < 0.05, |log2FC| > 1).

Diagnostic plots for the DE analysis

Diagnostic plots confirmed the validity of the analysis. The p-value distribution showed substantial enrichment for low p-values, indicating genuine differential expression rather than an all-noise result.

Most significant genes exhibited moderate to high expression levels (log₁₀ baseMean between 5 and 7), confirming that differential expression was primarily detected in well-expressed transcripts with sufficient statistical power. Very lowly expressed genes contributed minimally, further supporting the robustness of the filtering and normalization procedures.

Significantly differentially expressed genes in liver
Cre-AAV and GFP control
(FDR < 0.05, |log2FC| > 1)
gene baseMean log2FoldChange padj
Rps26-ps1 18.84 -6.60 0.00
Vcp-rs 65.58 -2.35 0.00
Rpl18-ps2 30.06 -2.97 0.00
Gm8566 108.75 -2.14 0.00
Gm10073 33.39 -2.97 0.00
Gm10263 18.29 -4.10 0.00
Rpl28-ps1 58.47 -1.93 0.00
Gstm2-ps1 53.61 -2.97 0.00
Gm45855 32.41 -2.00 0.00
Gm4332 37.02 -2.56 0.00
Gm11478 50.89 -2.14 0.00
Glo1-ps 43.49 -2.06 0.00
Gm16437 14.46 -3.16 0.00
Gm13611 13.15 -4.32 0.00
Gm15163 39.33 -1.89 0.00
ENSMUSG00000143433 29.55 -2.66 0.00
Gm9816 17.24 -2.16 0.00
Gm13436 32.37 -2.70 0.00
Gm10221 38.13 -2.11 0.00
Gm4366 59.20 -1.37 0.00
Glns-ps1 29.01 -2.54 0.00
Ccdc148 79.22 1.29 0.00
Eef2-ps2 25.88 -2.81 0.00
mt-Ts1 8.98 -4.92 0.00
Gm12481 52.96 -6.89 0.00
Lrp1 8579.16 -1.08 0.00
Gm11964 10.61 -3.07 0.00
Cox5b-ps 37.20 -1.81 0.00
Ftl1-ps1 524.56 -1.42 0.00
Gm14586 43.58 -2.27 0.00
2810001A02Rik 224.90 1.23 0.00
Gm6180 6.73 -3.00 0.00
Gstp-ps 133.94 -2.17 0.00
Slc41a2 264.20 -1.91 0.00
Rgs4 54.71 1.19 0.00
Eno1b 38.52 -2.73 0.00
Mmp7 15.72 4.03 0.00
Gm8730 51.91 -2.43 0.00
Rpl7-ps7 12.77 -1.98 0.00
Rps18-ps5 46.29 -2.48 0.00
Gm9794 75.37 -1.43 0.00
Gm7600 16.38 -2.82 0.00
Gm7049 17.59 -2.20 0.00
Bnip3 12935.81 1.05 0.00
Gm6204 32.82 -2.68 0.00
Gm5805 13.88 -3.10 0.00
Mov10 216.39 -1.22 0.00
Arg2 13.53 2.31 0.00
Mif-ps4 9.50 -3.53 0.00
ENSMUSG00000140143 99.65 2.25 0.00
ENSMUSG00000121793 109.23 -1.72 0.00
ENSMUSG00000128633 94.62 -2.10 0.00
Gm10076 1830.86 1.10 0.00
Gm10874 21.94 -1.72 0.00
ENSMUSG00000139771 170.64 -1.66 0.00
Lncbate1 94.08 1.37 0.00
Gm5540 32.59 -2.30 0.00
Gm5456 15.84 -2.87 0.00
Gm4149 35.41 -2.08 0.00
Gm12350 33.93 -1.94 0.00
Gm12174 27.14 -2.06 0.00
Gm10232 15.91 -2.22 0.00
E030018B13Rik 64.34 4.89 0.00
Gm16573 624.05 1.50 0.00
Rps10-ps1 15.50 -3.15 0.00
Gm9755 21.25 -1.95 0.00
Gm2962 11.66 -2.48 0.00
Slc36a1 409.28 -1.03 0.00
Gm44170 66.88 -1.72 0.00
ENSMUSG00000135866 247.05 -2.93 0.00
Snhg3 225.84 1.56 0.00
Myorg 671.66 -1.07 0.00
Abca2 552.28 -1.25 0.00
Rab11b-ps2 4.83 -3.51 0.00
Gm10039 20.42 -1.90 0.00
Nhsl3 894.50 -1.10 0.00
Mrpl23-ps1 12.15 -2.06 0.00
Gm56915 65.26 1.33 0.00
Gm6652 275.97 -1.11 0.00
Cd83 93.84 1.45 0.00
C4b 16639.92 -1.04 0.00
ENSMUSG00000126438 14.80 -2.41 0.00
Gm6733 15.30 -2.05 0.00
Rpl17-ps3 57.80 -1.67 0.00
Mmp15 527.71 -1.23 0.00
Pgam1-ps2 24.37 -1.83 0.00
ENSMUSG00000121505 150.78 1.60 0.00
Gm10288 116.74 -1.48 0.00
Adrm1b 37.12 -1.71 0.01
Ptprf 2408.30 -1.08 0.01
Gm57111 135.57 1.68 0.01
Gm12918 87.62 -1.41 0.01
Rpl19-ps11 113.24 -1.12 0.01
Got2-ps1 61.42 -1.01 0.01
Gm4459 14.98 -2.13 0.01
ENSMUSG00000137723 178.93 1.39 0.01
Rps16-ps2 81.63 -1.49 0.01
Hspd1-ps3 59.73 -1.01 0.01
Gm12251 16.03 -1.87 0.01
ENSMUSG00000142392 13.48 6.79 0.01
Znf41-ps 5.56 17.34 0.01
Abca3 1162.54 -1.05 0.01
Gm5422 13.91 -2.38 0.01
Sntg1 9.52 2.94 0.01
Tpt1-ps3 147.96 -1.14 0.01
Gpi-ps 16.78 -2.09 0.01
Gm9616 14.66 -2.26 0.01
Rps3a2 79.40 -1.17 0.01
Rpl23a-ps3 34.34 -1.59 0.01
Col1a1 50.22 -1.25 0.01
Gm10819 7.05 -2.83 0.01
Gm9625 29.65 -1.56 0.01
Gm12751 42.57 1.40 0.01
Sap18b 38.87 -1.28 0.01
Rps15-ps2 8.18 -2.50 0.01
Gm28857 68.21 2.11 0.01
Hspe1-rs1 39.18 -1.58 0.01
Rpl35rt 122.60 -1.57 0.01
Mfhas1 318.41 -1.03 0.01
Fn1 19990.89 -1.11 0.01
E230001N04Rik 24.89 2.48 0.01

To focus on high-confidence, biologically interpretable results, we applied an additional filtering step to remove pseudogenes (-ps, -rs), unannotated loci (Gm-, Rik, ENSMUS), mitochondrial (-mt), and ribosomal (-Rp) genes. This yielded 26 high-confidence nuclear genes that were consistently dysregulated in response to Cre-mediated ApoE4→ApoE2 conversion in liver, after adjusting for both sex and batch effects.

gene baseMean log2FoldChange padj
Ccdc148 79.22 1.29 0.00
Lrp1 8579.16 -1.08 0.00
Slc41a2 264.20 -1.91 0.00
Rgs4 54.71 1.19 0.00
Eno1b 38.52 -2.73 0.00
Mmp7 15.72 4.03 0.00
Bnip3 12935.81 1.05 0.00
Mov10 216.39 -1.22 0.00
Arg2 13.53 2.31 0.00
Lncbate1 94.08 1.37 0.00
Slc36a1 409.28 -1.03 0.00
Snhg3 225.84 1.56 0.00
Myorg 671.66 -1.07 0.00
Abca2 552.28 -1.25 0.00
Nhsl3 894.50 -1.10 0.00
Cd83 93.84 1.45 0.00
C4b 16639.92 -1.04 0.00
Mmp15 527.71 -1.23 0.00
Adrm1b 37.12 -1.71 0.01
Ptprf 2408.30 -1.08 0.01
Abca3 1162.54 -1.05 0.01
Sntg1 9.52 2.94 0.01
Col1a1 50.22 -1.25 0.01
Sap18b 38.87 -1.28 0.01
Mfhas1 318.41 -1.03 0.01
Fn1 19990.89 -1.11 0.01

6.3.2 Brain

In brain tissue from all mice, we identified 150 significantly differentially expressed genes between Cre-AAV and GFP control animals using the same statistical approach and thresholds (FDR < 0.05, |log2FC| > 1).

Diagnostic plots for the DE analysis

Diagnostic plots revealed that the p-value distribution from the Wald test was approximately uniform with no substantial enrichment of low p-values, suggesting a more modest transcriptional response in brain compared to liver. Nevertheless, the log₁₀ baseMean distribution for significant genes displayed moderate to high expression levels, indicating that detected changes occurred in well-expressed genes.

Significantly differentially expressed genes in brain
Cre-AAV and GFP control
(FDR < 0.05, |log2FC| > 1)
gene baseMean log2FoldChange padj
Gm15427 62.49 -2.26 0.00
Rpl35rt 127.75 -2.50 0.00
Rpsa-ps10 72.78 -2.05 0.00
Cox5b-ps 35.24 -2.07 0.00
Gm10288 97.74 -1.96 0.00
Rps16-ps2 55.97 -2.35 0.00
Rps6-ps4 115.70 -1.24 0.00
Rps19-ps6 38.23 -2.75 0.00
Gm4149 31.98 -2.53 0.00
Rps26-ps1 14.16 -5.64 0.00
Rps3a2 68.28 -1.92 0.00
4632415L05Rik 14.57 -5.24 0.00
Rpl19-ps11 99.78 -1.79 0.00
Gm10073 41.80 -2.71 0.00
Rpl17-ps10 78.73 -1.81 0.00
Gm6560 72.20 -1.37 0.00
ENSMUSG00000136525 49795.59 0.84 0.00
Gm11478 43.76 -2.41 0.00
Rps18-ps5 21.26 -3.69 0.00
Rpl14-ps1 56.17 -1.53 0.00
Rpl17-ps3 50.93 -2.15 0.00
ENSMUSG00000121793 81.80 -2.33 0.00
Eno1b 46.81 -2.12 0.00
Gm13456 187.81 -1.36 0.00
Rpl3-ps1 182.80 -1.08 0.00
Rps18-ps6 37.32 -2.34 0.00
Rpl23a-ps3 35.90 -1.85 0.00
Gm4332 27.77 -2.94 0.00
Rps13-ps1 30.42 -2.42 0.00
Gm9385 45.24 -1.82 0.00
ENSMUSG00000121779 68.48 -1.07 0.00
Gm13436 23.64 -3.16 0.00
Uba52rt 148.07 -1.35 0.00
Gm13611 9.55 -4.67 0.00
ENSMUSG00000143433 23.47 -2.86 0.00
Gm9794 52.50 -1.63 0.00
Gm14586 31.13 -2.25 0.00
ENSMUSG00000132604 24.15 -2.24 0.00
Rps23-ps1 47.48 -1.86 0.00
Rps25-ps1 12.78 -3.35 0.00
Gm7600 16.89 -2.87 0.00
Rpl9-ps6 211.35 -1.41 0.00
Gm11223 38.17 -1.56 0.00
Gm8730 21.73 -2.83 0.00
Gm8203 15.74 -2.36 0.00
Rps15a-ps7 15.33 -2.27 0.00
Gm7308 39.44 -1.47 0.00
Gm7536 94.83 -1.53 0.00
ENSMUSG00000121784 146.55 -1.04 0.00
Oaz1-ps 63.97 -1.58 0.00
Gm8566 20.17 -2.14 0.00
Gm9493 31.73 -1.93 0.00
Rplp2-ps1 22.23 -3.29 0.00
Gm4366 82.42 -1.31 0.00
Gm6204 28.49 -2.70 0.00
Gm15772 27.58 -2.14 0.00
Gm5436 14.19 -2.57 0.00
Gm5805 12.05 -3.52 0.00
Gm37530 12.73 2.30 0.00
Fus 4486.34 0.34 0.00
Gm12892 88.89 -1.29 0.00
Rps27a-ps2 48.15 -1.41 0.00
Gm4202 64.89 -1.10 0.00
Ndufs7 1547.25 -0.45 0.00
Ndufa5 1405.95 -0.36 0.00
Gm10263 12.36 -3.43 0.00
Gan 640.17 0.32 0.00
Ftl1-ps1 45.47 -1.53 0.00
Gm6710 221.28 -0.65 0.00
ENSMUSG00000140457 486.34 0.78 0.00
Gm8355 59.14 -1.07 0.00
Rps7-ps3 49.67 -1.95 0.00
Rpl18-ps2 19.31 -2.75 0.00
Gm5499 26.90 -1.55 0.00
Gm6789 41.50 -1.14 0.00
Rpl36a-ps2 20.10 -1.81 0.00
Aldoart1 15.29 -2.54 0.00
Rpl35a-ps3 34.77 20.50 0.00
ENSMUSG00000139771 140.98 -1.92 0.00
Rpl10a-ps1 65.56 -1.00 0.00
Rps10-ps2 42.98 -1.33 0.00
Gm8226 79.61 -1.11 0.00
Pgam1-ps2 69.70 -1.62 0.00
Ap2m1-ps 153.16 -0.79 0.00
Gstp-ps 13.32 -2.30 0.00
Dock9 3143.58 0.27 0.00
Rpl37rt 29.46 -5.75 0.00
Endog 301.92 -0.52 0.00
Nrtn 72.99 -0.92 0.00
1810049J17Rik 13.61 -3.05 0.00
Zmat5 271.78 -0.43 0.00
Rpl28-ps1 39.16 -1.84 0.00
Dido1 1334.11 0.26 0.00
Gm16437 13.68 -2.18 0.00
Gm6863 45.91 -1.32 0.00
Rps24-ps3 27.59 -2.13 0.00
Fkbp1b 449.39 -0.29 0.00
Olig1 2326.62 -0.35 0.00
Tnrc6c 1786.75 0.37 0.00
Gm6180 27.34 -1.48 0.00
Gm12174 18.35 -2.03 0.00
Ube2n-ps1 15.64 -1.65 0.00
Mif-ps4 10.48 -3.36 0.00
Gm14121 10.89 -2.17 0.00
ENSMUSG00000121655 13.32 -1.95 0.00
Pcsk1n 9849.26 -0.53 0.00
Gm6170 4.66 -3.46 0.00
Tmem223 586.05 -0.42 0.00
ENSMUSG00000121777 299.91 -0.97 0.00
Kmt2d 1891.12 0.41 0.00
4933431K14Rik 67.96 -0.80 0.00
Srrm4 875.61 0.33 0.00
Gm12254 19.49 -1.69 0.00
Bicral 1162.75 0.25 0.00
Gm10052 37.93 -0.98 0.00
Gpi-ps 58.96 -1.27 0.00
Rps10-ps1 13.08 -2.74 0.00
ENSMUSG00000121783 447.07 -0.85 0.00
Kmt2a 2860.62 0.33 0.00
Gria2 10476.71 0.27 0.00
Gm9816 29.45 -1.56 0.00
Pigyl 436.88 -0.44 0.00
Gm3571 7.47 -3.33 0.01
Rbm25 2275.66 0.36 0.01
Lars2 32337.98 0.83 0.01
Cdr1 119.55 1.24 0.01
Dohh 1074.71 -0.41 0.01
ENSMUSG00000128633 85.79 -2.98 0.01
Gm10250 42.45 -1.38 0.01
Gm10221 21.83 -1.99 0.01
Wdr89 111.21 -0.66 0.01
Rpl31-ps8 53.56 -1.35 0.01
Nudc-ps1 12.73 -1.88 0.01
Ramp1 1295.27 -0.36 0.01
Sap18b 30.52 -1.37 0.01
Gm4459 11.16 -2.27 0.01
Hspe1-rs1 14.02 -2.19 0.01
Atp6v0c-ps2 39.64 -1.13 0.01
ENSMUSG00000137224 84.10 -1.16 0.01
Gfra4 735.38 -0.30 0.01
Stmn3 10342.08 -0.32 0.01
Mt3 4865.41 -0.38 0.01
Nopchap1 1134.99 0.31 0.01
P3h4 305.18 -0.42 0.01
Rftn2 724.62 -0.28 0.01
B020010K11Rik 32.09 -2.03 0.01
Tnrc6b 2113.46 0.36 0.01
Mrpl12 1167.93 -0.42 0.01
Pdf 375.24 -0.41 0.01
Fbxl15 321.95 -0.53 0.01

After applying the same filtering criteria to exclude pseudogenes, unannotated loci, mitochondrial genes, and ribosomal protein genes, we retained 5 high-confidence nuclear genes that showed consistent dysregulation in brain tissue following Cre-mediated ApoE4→ApoE2 conversion, adjusted for batch and sex.

gene baseMean log2FoldChange padj
Eno1b 46.81 -2.12 0.00
Uba52rt 148.07 -1.35 0.00
Aldoart1 15.29 -2.54 0.00
Cdr1 119.55 1.24 0.01
Sap18b 30.52 -1.37 0.01

6.3.3 Common DEGs in liver and brain

To identify core molecular responses to ApoE isoform conversion that occur independently of tissue context, we intersected the lists of high-confidence DEGs from liver and brain. In the batch-adjusted analysis, only 2 genes were significantly differentially expressed in both tissues:

Eno1b, Sap18b