uap – Robust, Consistent, and Reproducible Data Analysis¶
uap executes, controls and keeps track of the analysis of large data sets. It enables users to perform robust, consistent, and reproducible data analysis. uap encapsulates the usage of (bioinformatic) tools and handles data flow and processing during an analysis. Users can use predefined or self-made analysis steps to create custom analysis. Analysis steps encapsulate best practice usages for bioinformatic software tools. Although uap focuses on the analysis of high-throughput sequencing data it can be extended to handle any analysis. But its plugin architecture allows users to add functionality. This would enable any kind of large data analysis.
uap is a command-line tool, implemented in Python, and runs under GNU/Linux. It takes a user-defined configuration file, which describes the analysis, as input. uap interacts with the analysis via subcommands.
The uap installation does not include all necessary tools for the data analysis. It expects that the required tools are already installed.
The recommended workflow to analyse data with uap is:
- Install uap (see Installation of uap)
- Optionally: Extend uap by adding new steps (see Extending uap Functionality)
- Write a configuration file to setup the analysis (see Configuration File)
- Start the analysis locally (see run-locally) or submit it to the cluster (see submit-to-cluster)
- Follow the progress of the analysis (see status)
- Optionally: share your extensions with others (send a pull request via github)
A finished analysis leaves the user with:
- The original input files (which are, of course, left untouched).
- The experiment-specific configuration file (see Configuration File). You should keep this configuration file for later reference and you could even make it publicly available along with your input files for anybody to re-run the entire data analysis or parts thereof.
- The result files and comprehensive annotations of the analysis (see Annotation Files). These files are located at the destination path defined in the configuration file.
Robust Data Analysis:¶
- Data is processed in temporary location. If and only if ALL involved processes exit gracefully, the output files are copied to the final output directory.
- Processing can be aborted and continued from the command line at any time. Failures during data processing do not lead to unstable state of analysis.
- The final output directory names are suffixed with a hashtag which is based on the commands executed to generate the output data. Data is not easily overwritten and this helps to check for necessary recomputations.
- Fail fast, fail often. During initiation uap controls the availability of all required tools, calculates the paths of all files of the analysis, and controls if these files exist. If any problems are encountered the user is requested to fix them.
- Steps and files are defined in a directed acyclic graph (DAG). The DAG defines dependencies between in- and output files.
- Prior to any execution the dependencies between files are calculated. If a file is newer or an option for a calculation has changed all dependent files are marked for recalculation.
- Comprehensive annotations are written to the output directories. They allow for later investigation of errors or review of executed commands. They contain also versions of used tool, required runtime, memory and CPU usage, etc.
- Single configuration file describes entire data analysis work-flow.
- Command-line tool to interacts with uap. It can be used to execute, monitor, and analyse defined work-flows.
Table of contents¶
- Installation of uap
- How-To Use uap
- Configuration File
- Command-Line Usage of uap
- Extending uap Functionality
- Software Design
- Annotation Files
- Available steps
- Source steps
- Processing steps
- API documentation