HP-GAS: prediction of Human Protein protein interactions based on automatic feature engineering and Genetic Algorithm driven Stacking method

HP-GAS is a software for prediction of human protein protein interactions based on graph, evolutionary and sequence features, and automatic feature engineering which utilizes genetic algorithm (GA) and automatic correlation based selection. HP-GAS uses the ensemble of models generated by machine learning (ML) algorithms as a method for PPI prediction, where automatic ensembling of ML algorithms was driven by supervized GA and unsupervized correlation filtering.

HP-GAS software was written in JAVA language and is available as standalone application, which can be executed on any operating system containing Java Virtual Machine. Minimum system requirements for HP-GAS are: RAM 1 GB; Disk space 1 GB.

In order to run the HP-GAS program it is necessary to install Java Runtime Environment 8 (JRE), which can be found for Windows, Linux, Mac OS and Solaris systems at: Java SE Runtime Environment 8 - Downloads

Please read the documentation for detailed information about the HP-GAS software and it's usage.

HP-GAS is a free software released under Apache License, Version 2.0.

HP-GAS application with required files and documentation is provided bellow.

BinariesHP-GAS_Binaries.zip697 MB79
DocumentationHP-GAS_Manual.pdf297 KB206
SequencesHP-GAS_Sequences.zip5.74 MB16
DatasetsHP-GAS_Datasets.zip76.75 MB28
Supplementary dataHP-GAS_Supplements.zip2.24 MB47

The file contains 15,650 human sequences, with UniProt identifiers and entrynames in FASTA format, for which the predictions can be calculated.

If using HP-GAS, please cite:
Sumonja N, Gemovic B, Veljkovic N and Perovic V. (2019) Automated feature engineering improves prediction of protein-protein interactions. In submission.