I'm encountering an issue with the MOSCOT library when generating Sankey diagrams to visualize cell lineage transitions for a single-cell RNA-seq dataset. The dataset contains three genotypes (WT, RunX1ko, RunX1/2Ko) with three time points (3, 7, and 28 days). I'm using moscot.problems.time.TemporalProblem to compute transport maps and a custom function to visualize the results as Sankey diagrams for each genotype (combined, WT, RunX1ko).
The Sankey diagrams for the combined dataset and WT genotype are satisfactory, showing specific lineage transitions between clusters across time points. However, for the RunX1ko genotype, the Sankey diagram exhibits proportional splitting: cells from each cluster at an earlier time point (e.g., 7 days) are distributed to all clusters at the next time point (e.g., 28 days) in a ratio proportional to the cell number ratio among clusters in the target time point. This results in a diffuse transport map that doesn’t reflect biologically meaningful lineage transitions.
Dataset Details
- Data: Single-cell RNA-seq AnnData object with genotypes (WT, RunX1ko, RunX1/2Ko).
- Time Points: 3, 7, and 28 days.
- Cell Count Imbalance: For RunX1ko, the 7-day time point has more cells than the 28-day time point (e.g., 1000 vs. 300 cells), unlike WT, which shows increasing cell counts over time.
Any guidance on optimizing MOSCOT for datasets with cell count imbalances or genotype-specific dynamics would be greatly appreciated.
I'm encountering an issue with the MOSCOT library when generating Sankey diagrams to visualize cell lineage transitions for a single-cell RNA-seq dataset. The dataset contains three genotypes (WT, RunX1ko, RunX1/2Ko) with three time points (3, 7, and 28 days). I'm using
moscot.problems.time.TemporalProblemto compute transport maps and a custom function to visualize the results as Sankey diagrams for each genotype (combined, WT, RunX1ko).The Sankey diagrams for the combined dataset and WT genotype are satisfactory, showing specific lineage transitions between clusters across time points. However, for the RunX1ko genotype, the Sankey diagram exhibits proportional splitting: cells from each cluster at an earlier time point (e.g., 7 days) are distributed to all clusters at the next time point (e.g., 28 days) in a ratio proportional to the cell number ratio among clusters in the target time point. This results in a diffuse transport map that doesn’t reflect biologically meaningful lineage transitions.
Dataset Details
Any guidance on optimizing MOSCOT for datasets with cell count imbalances or genotype-specific dynamics would be greatly appreciated.