Abstract
Combining data from two compatible studies for meta-analysis, we ran a series of exploratory DESeq2 runs, using variations of contrasts and sample sets. We found that differential expression of genes between pre-identified tolerant and sensitive samples was relatively independent of treatment (controls vs treated samples), indicating a possible innate DEG profile for tolerance.
Discussion
Among the five studies that we’re currently working with, Study 1 and Study 5 are the most compatible in terms of experimental design. Both use pre-identified tolerant and sensitive families, have Control and Treatment groups, sample gill tissue, and use similar preparation and analysis methodology. So at this juncture we chose to pursue our meta-analysis route with this more constrained set. Each of the two studies included samples taken at Day 7 after infection, and that seemed to be a point where differential expression was well manifested, so we used only Day 7 samples from each study.
We were working intially with the notion that the presence of Control and Treatment groups was important, allowing us to examine primarly the differences between Controls and treated samples across studies.
We ran a series of [nf-core/differentialabundance] runs for exploration.
However, intial contrasts for Treatment showed only one significantly differentially expressed gene! A second run, using only treated samples and contrasting for Condition also showed only one DEG. This led to a hunch that tolerance/sensitivity is shown more in innate differential expression (i.e. common to control and treated samples) between families, rather than in response to treatment.
Further differentialabundance runs, first on each of the two studies separately, and then again on the combined set from both studies, gives a promising indication that the differential abundance between sensitive and tolerant groups is exhibited in both control and treated samples.
Details
All runs were performed on the Seqera platform.
Seqera run condescending_davinci_5 (config) combined Study 1 and Study 5 samples and contrasted for Treatment.
Sample sheet
Contrasts file Contrast on Treatment, block for batch.
Matrix
Results are in s3://steveyost-seqera/Cvirg_Pmarinus/deseq2/study1and5_D7_results
PCA showed separation in Condition in PC2, little separation in Treatment, and some in Batch in PC3 and PC4.
Only two genes were significantly up-regulated for Treatment; none down-regulated.
This led to the hunch that tolerance/sensitivity is shown more in innate expression rather than response to treatment (infection).
Seqera run disturbed_fermat_2 (config) combined Study 1 and Study 5 samples, but only treated samples (no controls) and contrasted for Condition.
Sample sheet
Contrasts file Contrast on condition, block for batch.
Matrix
Results are in s3://steveyost-seqera/Cvirg_Pmarinus/deseq2/study1and5_D7_no_controls_results
Only one gene was found to be differentially expressed for Condition among these treated samples.
Seqera run crazy_newton (config) used Study 5 only, including Control and Treatment samples, but contrasting only on Condition, with no blocking.
Sample sheet
Contrasts file
Results are in s3://steveyost-seqera/Cvirg_Pmarinus/deseq2/study5_D7_condition_results
PCA showed clear separation between Condition tolerant and sensitive in PC1, some separation of Batch in PC2 and little separation of Treatment.
1158 genes were significantly up-regulated for Condition, 1099 down-regulated.
Seqera run mad_mayer_3 (config) is the same as crazy_newton, but for Study 1 only.
PCA showed strong separation in Treatment in PC1 and PC2, and good separation in Condition in PC2 and PC3.
1448 genes were significantly up-regulated for Condition, 1313 down-regulated.
Finally, Seqera run Issue_44_study1and5_condition (config) used all samples (Control and Treatment) from Study 1 and Study 5, contrasting on Condition, blocking for Batch.
Sample sheet ( same as condescending_davinci_5)
Contrasts file (Same as disturbed_fermat_2)
Note it used filtering_min_samples: 10
Results are in s3://steveyost-seqera/Cvirg_Pmarinus/deseq2/Issue_44_study1and5_D7_condition_results
PCA showed strong batch separation in PC1 (21.3%) (note this is pre-batch-correction in the pipeline), separation for Treatment in PC1, and separation for Condition in PC3 (5.5%)
1046 genes were significantly up-regulated for Condition, 905 down-regulated.
Download the report HTML page.