Pipeline Manual: Trajectory Analysis across Cortical Layers

Run the Pipeline in One Command

To run the pipeline, configure the script run_analysis.sh first. Things to configure:

• data_prefix: the data prefix of the dataset (e.g. Kriegstein)
• run_path: the present working directory that contains these codes
• data_path: the directory to the datasets that are required in this analysis. For now it is the Hybrid_cell_scheme directory.
• use_coding: boolean value to decide whether or not only the coding features are used in the analysis. Default is FALSE meaning that all features are included.
• run_parallel: boolean value to decide whether or not to run fitGAM in parallel. If we want to complete the analysis in a single command, we need to set it to FALSE, but it might take longer time.

After configuring these parameters, simply submit the job to Slurm, which will complete the trajectory analysis and fitGAM analysis. The Seurat objects are GAM objects will be saved in ./data, and all the results will be saved in ./results.

sbatch ./run_analysis.sh

After the pipeline is finished, run the 4_postprocessing_dsanalysis.R interactively to produce the smoothly varying genes, heatmap visualization and common genes across datasets. Note that we will need to specify which datasets are to be included.

Run the fitGAM in Parallel (optional)

Some datasets are large with many cells and features (e.g. Ma_Sestan, UCLA-ASD). Set run_parallel=TRUE would utilize the dSQ job arrays system to run fitGAM in parallel. Doing this requires additional steps but might save some running time.

Step 1

Set run_parallel=TRUE in run_analysis.sh and submit the job to Slurm.

sbatch ./run_analysis.sh

Step 2

When the dSQ jobs are finished running, submit an addtional script after_fitGAM_parallel.sh to Slurm to complete the pipeline. Make sure to keep the data_prefix and run_path consistent as in the run_analysis.sh script.

sbatch ./after_fitGAM_parallel.sh

Step 3

After the pipeline is finished, run 4_postprocessing_dsanalysis.R interactively to produce and visualize the results.

Credits to Tiernon Riesenmy (link), Dr. Cagatay Dursun and Dr. Prashant Emani.