xinluzhu/jump

Introduction

JUMP (JUmbo Mass spectrometry-based Proteomics tool) is a software suite for processing mass spectrometry-based proteomics data including peptide/protein identification, filtering of identified peptide-spectrum matches (PSMs), quantification of peptides/proteins and so forth.

Basic installation

A basic install is sufficient for multicore laptops, desktops, workstations and servers. For a basic install, use the bootstrap.sh script provided in the repo. Then,

1. Place the JUMP distribution source in the desired location (call this <path to JUMP>)
2. Change your working directory to the top level of the JUMP install and run bashbootstrap.sh

Obtaining JUMP source

You can obtain the latest version of JUMP from git; simple clone the git repository:

    git clone https://github.com/JUMPSuite/JUMP.git JUMP

in the directory where you would like JUMP to be installed (call this directory <path to JUMP>). Note that JUMP does not support out-of-place installs; the JUMP git repository is the entire installation. History of JUMP releases is provided by git tags.

Bootstrapping

To get dependencies installed, we recommend Conda, and we have provided a bootstrapping script bootstrap.sh that downloads all dependencies and installs them alongside JUMP. Execute

    ./bootstrap.sh

and it will create a new directory conda in the current working directory. The directory conda will contain the conda environment to be used by JUMP. JUMP will be set up to use the PERL and python interpreters in that environment.

Once bootstrap.sh is finished, the JUMP suite will be configured. JUMP components then are available via the executable script <path to JUMP>/JUMP/bin/jump. For convenience, you may simply append <path to JUMP>/JUMP/bin to your PATH environment variable as PATH=$PATH:<path to JUMP>/JUMP/bin. You will not need to activate the conda environment bootstrap.sh creates. You are now ready to run JUMP!

You may use existing installations of PERL, python and R for JUMP, but this is not recommended. PERL 6 is not supported.

How to run JUMP

Please see the link for JUMP manual. Sample datasets (mzXML files for TMT-11plex and 16plex), parameter files and database files can be downloaded at https://www.stjuderesearch.org/site/lab/peng

Installation on a HPC cluster

JUMP can be configured to interact with an HPC job manager on a compute cluster. Extra command-line arguments are necessary to bootstrap.sh. An example for an LSF cluster with no "service nodes" is

    ./bootstrap.sh CLUSTER=1 \
        DEFAULT_BATCH_CMD='bsub -K -P peng_grp -q standard' \
        COMPUTE_ON_LOGIN_NODE=0 \
	BATCH_DISPATCH_LAG=.01 \
	MAX_DISPATCH_WORKER_PROCS=128

The extra arguments to bootstrap.sh explained:

Argument	Meaning
CLUSTER	1 for cluster installation, 0 for local installation
DEFAULT_BATCH_CMD	The command used to submit scripts from the command line (see below for more information)
COMPUTE_ON_LOGIN_NODE	Set to 1 if your cluster has "service nodes" from which jobs are launched, 0 otherwise
BATCH_DISPATCH_LAG	Lag between batch job submissions to avoid overwhelming the job manager
MAX_DISPATCH_WORKER_PROCS	Throttling factor to prevent JUMP from spawning more than MAX_DISPATCH_WORKER_PROCS search jobs; useful for ensuring fairness in job priority or to prevent overuse of resources on service nodes

Batch command configuration

When CLUSTER=1, JUMP will use your job manager (e.g. Slurm, SGE, LSF) to submit and monitor jobs on your compute cluster. Your job manager must allow jobs defined as throwaway command line parameters. For example, you must be able to do

$ bsub -P peng_grp -q standard -K "perl myscript1.pl && perl myscript2.pl"

The command strings you create must create job submissions that block until the batch job completes. Use the following flags, based on your job manager:

Job Manager	Flag
SGE	-synch
LSF	-K
Slurm	-W

JUMP, batch commands, and "tool types"

The various stages of JUMP's pipeline have differing compute requirements; some require more memory, some may have longer wall clock times. You can create customized job submission strings for each of these "tool types" to allow for more memory or longer run times. If you do not create a customized job submission command for a tool type, JUMP will just use the DEFAULT_BATCH_CMD. The available "tool types" are:

Tool type	Explanation
JUMP_DATABASE_BATCH_CMD	Components used for database construction. Memory and wall clock time intensive.
JUMP_SEARCH_BATCH_CMD	Components used to run the job management and aggregation front-end of JUMP search. Memory intensive when COMPUTE_ON_LOGIN_NODE=0, these jobs must persist until the entire mzXML input is searched.
JUMP_QUANTIFICATION_BATCH_CMD	Components used for quantification. Memory intensive.
JUMP_FILTER_BATCH_CMD	Components used for filter. Very memory intensive.
JUMP_LOCALIZATION_BATCH_CMD	Components used for site localization. Very memory intensive.
RUNSEARCH_SHELL_BATCH_CMD	Sub-components used for searching individual spectra. Neither compute, memory or time intensive.