A computational method for the prediction of peptide hormones peptide sequences using an ensemble of machine learning and similarity-based methods

Peptide hormones serve as genome-encoded signal transduction molecules that play essential roles in multicellular organisms, and their dysregulation can lead to various health problems. In this study, we propose a method for predicting hormonal peptides with high accuracy. The dataset used for training, testing, and evaluating our models consisted of 1174 hormonal and 1174 non-hormonal peptide sequences. Initially, we developed similarity-based methods utilizing BLAST and MERCI software. Although these similarity-based methods provided a high probability of correct prediction, they had limitations, such as no hits or prediction of limited sequences. To overcome these limitations, we further developed machine and deep learning-based models. Our logistic regression-based model achieved a maximum AUROC of 0.93 with an accuracy of 86% on an independent/validation dataset. To harness the power of similarity-based and machine learning-based models, we developed an ensemble method that achieved an AUROC of 0.96 with an accuracy of 89.79% and a Matthews correlation coefficient (MCC) of 0.8 on the validation set. To facilitate researchers in predicting and designing hormone peptides, we developed a web-based server called HOPPred. This server offers a unique feature that allows the identification of hormone-associated motifs within hormone peptides. The server can be accessed at: https://webs.iiitd.edu.in/raghava/hoppred/.Please read/cite the content about the HOPPred for complete information including algorithm behind the approach.

Kaur, D., Arora, A., Vigneshwar, P., & Raghava, G. P. (2023). Prediction of peptide hormones using an ensemble of machine learning and similarity-based methods. bioRxiv, (), 2023.05.15.540764. Accessed April 25, 2024. https://doi.org/10.1101/2023.05.15.540764.

The standalone version is available to download at https://webs.iiitd.edu.in/raghava/hoppred/hoppred.zip The Standalone version of hoppred is written in python3 and following libraries are necessary for the successful run:

scikit-learn Pandas Numpy For the successful run of the standalone, please download and install the latest version for BLAST from https://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/LATEST/.

To know about the available option for the stanadlone, type the following command:

python3 hoppred.py -h

To run the example, type the following command:

python3 hoppred.py -i peptide.fa

This will predict if the submitted sequences are hormonal peptides or non-hormonal peptides. It will use other parameters by default. It will save the output in "outfile.csv" in CSV (comma seperated variables).

usage: hoppred.py [-h] -i INPUT [-o OUTPUT] [-t THRESHOLD] [-m {1,2}] [-d {1,2}]

Please provide following arguments

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
                        Input: protein or peptide sequence in FASTA format or
                        single sequence per line in single letter code
  -o OUTPUT, --output OUTPUT
                        Output: File for saving results by default outfile.csv
  -t THRESHOLD, --threshold THRESHOLD
                        Threshold: Value between 0 to 1 by default 0.5
  -m {1,2}, -- model Model
                        Model: 1: AAC based RF, 2: Hybrid, by default 1

Input File: It allows users to provide input in two formats; i) FASTA format (standard) (e.g. peptide.fa) and ii) Simple Format. In case of simple format, file should have one one peptide sequence in a single line in single letter code (eg. peptide.seq).

Output File: Program will save result in CSV format, in case user do not provide output file name, it will be stored in outfile.csv.

Threshold: User should provide threshold between 0 and 1, please note score is proportional to hormonal potential of peptide.

Models: In this program, two models have beeen incorporated: i) Model 1 for predicting given input peptide/protein sequence as Hormonal and Non-hormonal peptide/proteins using Logistic Regression based on top 50 features of the peptide/proteins;

ii) Model 2 for predicting given input peptide/protein sequence as Hormonal and non-hormonal peptide/proteins using Hybrid approach, which is the ensemble of Logistic Regression+ BLAST+ MERCI. It combines the scores generated from machine learning (LR), MERCI, and BLAST as Hybrid Score, and the prediction is based on Hybrid Score.

It contain following files, brief descript of these files given below

INSTALLATION : Installation instructions

LICENSE : License information

envfile : This file provide the path information for BLAST and MERCI commands,and data required to run BLAST and MERCI

Database: This folder contains the BLAST database and IgE motif files

progs : This folder contains the program to run MERCI

README.md : This file provide information about this package

hoppred.py : Main python program

rf_model : Model file required for running Machine-learning model

MERCI_motif_locator.pl : Perl script for locating motifs using MERCI

peptide.fa : Example file contain peptide sequences in FASTA format

peptide.seq : Example file contain peptide sequences in simple format

protein.fa : Example file contain protein sequences in FASTA format