GSE166504_experiment_card.md

analysis_date	h5ad_file	species	total_cells	design_type	edviz_grammar	factors	tool_version
2025-11-12	GSE166504.h5ad	Mouse (Mus musculus)	82192	Unbalanced factorial design with crossed factors	(CellFraction(2) × DietTimepoint(4)) > Sample(38) : CellType(13)	cell_fraction diet_timepoint sample cell_type	0.1.0

Experimental Design Card

Dataset Information

File: GSE166504.h5ad Analysis Date: 2025-11-12 Species: Mouse (Mus Musculus) Total Cells: 82,192

Experimental Context

Experiment Type: Liver cell profiling under different diet conditions

Research Question: How do high-fat high-sugar (HFHS) diet and duration affect liver cell populations in hepatocytes vs non-parenchymal cells?

Factor Descriptions:

• Cell Fraction: Two major liver cell fractions: Hepatocytes (parenchymal cells performing metabolic functions) vs NPC (non-parenchymal cells including immune, endothelial, and stellate cells)
• Diet Timepoint: Diet conditions and duration: 15weeks, 30weeks, 34weeks on HFHS diet, and Chow (control diet). Note: Hepatocyte fraction lacks 34weeks timepoint, creating an unbalanced design
• Sample: Individual samples representing technical replicates (Captures) from biological replicates (Animals) across cell fraction and diet/time combinations. Total of 38 samples: 16 Hepatocyte samples (across 3 timepoints) and 22 NPC samples (across 4 timepoints)
• Cell Type: Cell populations identified by marker expression: immune cells (B, T, DCs, pDCs, NK, Neutrophils), liver-specific cells (Hepatocytes, Kupffer cells, Stellate cells, Hepatic progenitor cells), stromal cells (Endothelial cells, Myofibroblasts), and monocyte/macrophages

Design Structure

Identified Factors

Factor	Levels	Type
Cell Fraction	2	Treatment
Diet Timepoint	4	Treatment
Sample	38	Replicate
Cell Type	13	Observation

Design Classification

This dataset exhibits a hierarchical nested design with multiple levels of nesting. Cell Types are observed across all samples, creating a crossed relationship with the nested structure.

Design Diagram

┌──────────────────── Design Structure ────────────────────┐
│                                                          │
│ CellFraction(2)  ────×──── DietTimepoint(4)              │
│    ↓                                                     │
│                                                          │
│ Sample(38)                                               │
│    :                                                     │
│                                                          │
│ CellType(13)                                             │
│                                                          │
└──────────────────────────────────────────────────────────┘

Grammar Notation

(CellFraction(2) × DietTimepoint(4)) > Sample(38) : CellType(13)

Distribution Summary

Samples per Cell Fraction: 16 - 22 (mean: 19)
Samples per Diet Timepoint: 2 - 12 (mean: 9)
Cells per Cell Type: 99 - 25,848 (mean: 6,322)

Analysis Considerations

This design structure has implications for statistical analysis:

Random Effects Modeling: The nesting of sample within cell_fraction indicates that sample-specific variation should be modeled as a random effect. When testing for cell_fraction effects, use mixed-effects models with random intercepts for sample (e.g., ~ cell_fraction + (1|sample) in lme4 notation).

Aggregation Strategy: For differential expression testing, pseudobulking to the sample level preserves the experimental unit structure. Aggregate cells to sample-by-cell_type pseudobulk profiles before applying standard DE methods, treating samples as biological replicates.

Contrast Specification: When comparing cell_fractions, ensure contrasts are computed at the sample level, not the cell level, to avoid pseudoreplication and inflated Type I error rates.

Design Notes

The design is unbalanced: Hepatocyte samples lack the 34weeks timepoint (only present in NPC samples)

Samples include both biological replicates (Animals) and technical replicates (Captures), with varying numbers of captures per animal