Administrator Manual General Information: Python 3.5:
This application is dependent upon several libraries for Python 3.5:
* pip==9.0.1
* Flask==0.12.2
* tqdm==4.17.1
* numpy==1.13.1
Which are specified in the requirements.txt document.
The application’s environment variables are set and initialized by app.py, including --entry-point, which points the server to the main program to execute.
Other configuration options are specified in skr_config.py, including constants.
Files run with and produced by the algorithm can be quite large. The following file size details should give an idea of minimum system requirements that should be met to run SEEKR effectively. Gencode human 27908 sequences 519,703 lines Sequences length: mean=1027, std=1867, median 636 longest=205012, #10=23112, #300 < 5000 File size – 8MB zipped; 30MB unzipped Gencode mice 16679 sequences 24MB unzipped Large test file 234,537 sequences 29,748,112 lines File size – 560MB zipped; 2GB unzipped Frequency counts 4^k * float32size * N_sequences Pearsons NNdouble_size 8 bytes Example: 5942MB for gencode pearsons as numpy array 410GB for large test file ~1.8MB for every row of matrix corr2 NMdouble_size Cache file sizes of the different kmer lengths
- 266,944
- 1,067,536
- 4,269,904
- 17,079,376
- 68,317,264
- 273,268,816
- 1,093,075,024
- 446,608
- 1,786,192
- 7,144,528
- 28,577,872
- 114,311,248
- 457,244,752
- 1,828,978,768
Create a local instance of the application on your machine by either cloning the git repository, or downloading it contents
Before spinning the local server up make sure to run precompute_sequence_sets.py. If this script is not run before initializing the application, errors will be encountered when using the standard gencode sets to run the algorithm.
Executing either app.py or seekrServer.py will boot the application locally.
This application is built to run on a VM instance of the Google Cloud Platform.
The first step to deployment is installing Openshift command line tool, instructions for which can be found here: https://help.unc.edu/help/carolina-cloudapps-installing-the-command-line-cli-tools/
Once the command line tool is installed, use it to establish an ssh pipeline between the server and the project’s git repository by following the instructions here: https://help.unc.edu/help/carolina-cloudapps-ssh-keys/
Once the ssh keys are created and saved, the OpenShift console allows you to build at will and will automatically deploy successful builds to the server.
Important: precompute_sequence_sets.py must be executed on the server before the application can be run correctly. This can be done from your local machine using the command line tool, os python3 precompute_sequence_sets.py or from the web console’s built in terminal.