LSPpred

Random Forest prediction model for leaderless secretory proteins, based on features from curated lists of positive training data.

Overview

LSPpred has two sub-modules:

• The main module LSPpred which is based on a curated list of likely unconventionally secreted proteins in Arabidopsis

• A secondary module SPLpred which is SecretomeP-like, in that it is based on classically secreted proteins with their signal peptide removed - notably from Arabidopsis, as opposed to the human data in the eukaryotic version of SecretomeP

Result of both modules are included in the .CSV output file

Each module has a built in cutoff to call a true-positive LSP with an estimated false-positive rate of 0.05.

The output file contains the following columns:

• Sequence - sequence ID
• LSPpred_probability - model predicated probability of LSP status
• LSPpred - TRUE if LSPpred_probability exceeds builtin cutoff
• SPLpred_probability - - model predicated probability of LSP status
• SPLpred - TRUE if SPLpred_probability exceeds builtin cutoff
• Either - TRUE if either of LSPpred and SPLpred are TRUE
• Consensus - TRUE if both of LSPpred and SPLpred are TRUE

Optionally the --low flag can be used to add low confidence results for each module, where LSPpred_probability or SPLpred_probability exceeds 0.5

Usage

usage: python lsppred.py [-h] [--version] [--low] [--output OUT_FILE]
                  [--log LOG_FILE]
                  FASTA_FILE

Predict LSPs in a FASTA file

positional arguments:
  FASTA_FILE         Input FASTA protein file

optional arguments:
  -h, --help         show this help message and exit
  --version          show program's version number and exit
  --low              Add low confidence predictions
  --output OUT_FILE  save output in CSV format to OUT_FILE
  --log LOG_FILE     record program progress in LOG_FILE

Installation

LSPpred can be run from the command line.

Web server

A webserver is available at: http://lsppred.lspdb.org/

Requirements

LSPpred requires the followinf packages to be isntalled, as outlined in requiremetns.txt:

• tabulate
• pandas
• biopython
• scikit-learn==0.21.2
• imbalanced-learn==0.5.0

Additionally PROfet by Ofer and Linial is required to be available in the src directory. As a convience, the required files are included in this repository.

Ofer, Dan, and Michal Linial. "ProFET: Feature engineering captures high-level protein functions." Bioinformatics (2015): btv345.

Docker

Citation

LSPpred suite: tools for leaderless secretory protein prediction in plants

Andrew Lonsdale, Laura Ceballos-Laita , Daisuke Takahashi, Matsuo Uemura, Javier Abadía, Melissa J. Davis, Antony Bacic and Monika S. Doblin*. 

In Preparation.