Cross Dataset Comparison

Cross-dataset comparison

Principle

Each dataset is filtered, normalized and tested within itself. Raw or normalized values are never merged, pooled or renormalized across datasets, and no p-value is computed on expression pooled across datasets. Comparisons operate on within-dataset statistics of matched genes. Merging studies would require assumptions about batch structure that the tool does not make.

Adding datasets

Up to eight datasets can be added to the session. Each has its own optional design file and species setting.

Gene matching

Situation	Matching
Same species	Case-normalized gene symbol
Mouse and human	MGI strict one-to-one orthologs
Custom ID map loaded	The map, applied before the rules above

Identifiers are decoded with each dataset's own species annotation. If matching yields few genes, the slot reports the identifier types detected on each side and the remedy.

Pairwise comparison

Output	Statistic
Fold-change scatter	log₂FC in X against log₂FC in Y, coloured by quadrant; Pearson r
t concordance	Pearson correlation of moderated t-statistics
Hit-list overlap	One-sided hypergeometric (Fisher) test of shared up and shared down hits
RRHO map	Rank–rank hypergeometric overlap, computed as in the Bioconductor RRHO package; under-enrichment shown as a signed extension
Bland–Altman plot	Mean against difference of log₂FC
Shared and discordant lists	Genes significant in both datasets, ranked

All datasets

Compute across all datasets produces the correlation matrix and a consensus table. The forest plot shows one gene's log₂FC with confidence intervals in every dataset.

Signature transfer

A module from dataset A (up- or down-regulated DE genes, or a Discovery module) is scored in each other dataset as a mean z-score. Discrimination is summarized as the Mann–Whitney AUC between that dataset's groups and calibrated against 200 seeded size-matched random signatures scored identically.

References

Plaisier SB et al. (2010) Nucleic Acids Res 38:e169. Wu D and Smyth GK (2012) Nucleic Acids Res 40:e133.