- Electron Cross-Platformed
BioSelector is a cross-platform application powered by machine learning that provides Ensemble and other Feature Selection techniques for biomarker discovery. Even though different domain problems could benefit from it if the dataset model fits the required one, we still didn't investigated its potential beyond gene expression data.
The application is part of a scientific study that investigated Ensemble Feature Selection for biomarker discovery out of RNA-seq and Microarray datasets provided by TCGA and ArrayExpress. If you have scientific interests or want to use our package in formal reports, we kindly ask you to cite us in your publication: Colombelli, F., Kowalski, T.W. and Recamonde-Mendoza, M., 2022. A hybrid ensemble feature selection design for candidate biomarkers discovery from transcriptome profiles. Knowledge-Based Systems, 254, p.109655.
This work was developed at the Institute of Informatics, Universidade Federal do Rio Grande do Sul and Bioinformatics Core, Hospital de Clínicas de Porto Alegre.
Right now, the application is at an early stage development and its MVP was only executed on Linux systems. The application was built using Electron backended by Python/R/Rcpp and frontended with React framework utilizing a template provided freely by Creative Tim.
After more testing, this README will be updated with a detailed usage and installation guide.
Due to its early stage development phase, BioSelector still does not have production ready packages with installable binaries, thus, its usage is limited to the development version, and the installation is more manual and prone to errors.
Start by following the installation steps for efs-assembler, the backend solution of BioSelector. After that, install the latest stable version of Node.js (we have tested it on Node.js v10.24.1, v14.17.3 and v14.24.4).
Then, clone the repository and access its directory:
$ git clone https://github.com/colombelli/bioselector.git
$ cd bioselector
On Linux systems run the installation node script:
$ npm run install:clean
On Windows and macOS systems, delete the node_modules folder (if present from previous installation attempts) and run the node installation command:
$ npm install
Change the directory to bioselector and run the package.json script start:
$ cd bioselector
$ npm run start
- For information about the supported datasets format and other charcteristics, read this
- For information about the results folder structure and explanation about the generated files and folders, read this
- For information about the constraints surrounding the addition of new feature selection algorithms, read this
- For information about the constraints surrounding the addition of new aggregation algorithms, read this
@article{COLOMBELLI2022109655,
title = {A hybrid ensemble feature selection design for candidate biomarkers discovery from transcriptome profiles},
journal = {Knowledge-Based Systems},
volume = {254},
pages = {109655},
year = {2022},
issn = {0950-7051},
doi = {https://doi.org/10.1016/j.knosys.2022.109655},
url = {https://www.sciencedirect.com/science/article/pii/S0950705122008383},
author = {Felipe Colombelli and Thayne Woycinck Kowalski and Mariana Recamonde-Mendoza},
keywords = {Feature selection, Ensemble learning, Biomarkers discovery, Microarray, Bioinformatics, High-dimensional data}
}
This project was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001, Conselho Nacional de Desenvolvimento Científico e Tecnológico (project CNPq/AWS 032/2019, process no. 440005/2020-5), and Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul (FAPERGS).