How marionilab/emptydrops2017 works
Main Components
• Data Downloading and Preparation: Represented by the "data" folder, which contains shell scripts (download_10x.sh, download_placenta.sh) for acquiring raw datasets from external sources.
• Analysis Pipelines: Encapsulated in the "analysis" directory. Each subfolder (neurons_900, pbmc4k, placenta) contains its own self-contained analysis (analysis.Rmd) and plotting module (plot_maker.R). Treat these as separate processing units that take raw datasets and generate results.
• Simulation Framework: Located in the "simulations" directory with scripts (simrun.R, plotsim.R, functions.R) that simulate data scenarios to test cell detection methods.
• Real Data Analysis: Found in the "real" directory with R scripts (realrun.R, negcheck.R) that tackle the analysis of actual datasets and validate the statistical properties (like p-value distributions).
• Manuscript Compilation: The "manuscript" directory includes LaTeX files, a Makefile for compiling the document, and supporting files (bibliography, class files, figures). This component assembles figures and tables generated by the analysis into a final paper.
Relationships and Interactions
• Data flows from the Data Downloading scripts to the Analysis and Real Data Analysis pipelines. Both analysis streams consume external datasets.
• The Simulation component stands relatively parallel to the main analysis, providing an independent validation of cell detection methods.
Architectural Patterns and Technologies
• Pipeline/Workflow organization: Each major task (data acquisition, simulation, analysis, manuscript generation) is neatly separated, making it a modular structure.
• Scripting Nodes: Most of the functionality is driven by R and R Markdown (for reproducible analysis and plot generation), along with shell scripts for data download and a Makefile for LaTeX compilation.
• Emphasis on Reproducibility: The separation of concerns (individual directories per dataset and analysis type) emphasizes modular and reproducible research practices.
• Three blocks under “Analysis” (one for each data set: neurons_900, pbmc4k, placenta) showing the paired scripts for analysis and plotting.
• A “Real Data Analysis” block that shows scripts performing the application of cell detection methods.
• A “Simulations” block that captures simulation code and plotting for evaluating methods.
• Finally, all outputs converge into a “Manuscript Generation” block, highlighting the integration of results (figures, tables) into the final paper.
• Annotations that describe the directional flows and dependencies between these components.