Back to Search Start Over

An evolutionary approach to the discretization of gene expression profiles to predict the severity of COVID-19

Authors :
Mouhrim, Nisrine
Tonda, Alberto
Rodríguez-Guerra, Itzel
Kraneveld, Aletta D.
Rincon, Alejandro Lopez
Fieldsend, Jonathan E.
Afd Pharmacology
Pharmacology
Source :
GECCO '22: Proceedings of the Genetic and Evolutionary Computation Conference Companion, 731. Association for Computing Machinery, STARTPAGE=731;TITLE=GECCO '22: Proceedings of the Genetic and Evolutionary Computation Conference Companion
Publication Year :
2022

Abstract

In this work, we propose to use a state-of-the-art evolutionary algorithm to set the discretization thresholds for gene expression profiles, using feedback from a classifier in order to maximize the accuracy of the predictions based on the discretized gene expression levels, while at the same time minimizing the number of different profiles obtained, to ease the understanding of the expert. The methodology is applied to a dataset containing COVID-19 patients that developed either mild or severe symptoms. The results show that the evolutionary approach performs better than a traditional discretization based on statistical analysis, and that it does preserve the sense-making necessary for practitioners to trust the results.

Details

Language :
English
Database :
OpenAIRE
Journal :
GECCO '22: Proceedings of the Genetic and Evolutionary Computation Conference Companion, 731. Association for Computing Machinery, STARTPAGE=731;TITLE=GECCO '22: Proceedings of the Genetic and Evolutionary Computation Conference Companion
Accession number :
edsair.doi.dedup.....4c24580b95e5344031dd2c49a540d355