We showcase moslin on C. elegans embryogenesis in Fig. 3 and corresponding supplementary Figures. The following table describes the files that need to be run to reproduce our analysis. Scripts and notebooks should be executed in the order given by the table. Data files for this analysis are available from figshare, details are given in the notebooks themselves.
| File/Folder | What it does |
|---|---|
run_gridsearch |
Benchmark moslin, pure Gromow-Wasserstein, pure OT and LineageOT across a grid of hyperparameters using wandb. The gridsearch parameters are specified in moslin.yml and lot.yml for moslin and LineageOT, respectively. |
ML_2024-06-05_prepare_c_elegans.ipynb |
Clean up provided metadata, add TF annotations, aggregate clusters and assign custom colors. |
ML_2024-06-08_explore_c_elegans.ipynb |
Visualize the two data subsets employed in the manuscript: all cells with precise lineage information and the ABpxp lineage. |
ML_2024-06-04_visualize_gridsearch_results.ipynb |
Visualize the benchmark results in terms of the mean error. |
ML_2024-03-12_GW_performance.ipynb |
Show that pure GW performs better on precise lineage information when initialized with the OT solution. |
ML_2024-07-08_cost_matrix_permutations.ipynb |
Demonstrate the effect of perturbing cost matrix elements. |
ML_2024-03-11_compute_couplings.ipynb |
Compute couplings using hyperparmeters identified in the gridsearch. |
ML_2024-03-12_palantir.ipynb |
Run Palantir and MAGIC to obtain a pseudotime and imputed gene expression, respectively. Imputed data is only used to visualize gene expression trends. |
ML_2024-03-12_zoom_in_abpxp.ipynb |
Zoom-in onto one pair of time points and explore method differences. |
ML_{DATE}_cellrank2_{METHOD}.ipynb |
Compare cellular trajectories by combining CellRank 2 with different methods (METHOD in {ot, gw, moslin, LineageOT, palantir, cytotrace}) |
ML_2024-06-29_terminal_state_robustness.ipynb |
Investigate robustness when changing the number of terminal states. |