An Innovative Perspective on Metabolomics Data Analysis in Biomedical Research Using Concept Drift Detection

SCHWARZEROVÁ, Jana, Adam BAJGER, I. PIERDOU, Lubomír POPELÍNSKÝ, Karel SEDLÁŘ a W. WECKWERTH. An Innovative Perspective on Metabolomics Data Analysis in Biomedical Research Using Concept Drift Detection. Online. In Yufei Huang and Lukasz A. Kurgan and Feng Luo and Xiaohua Hu and Yidong Chen and Edward R. Dougherty and Andrzej Kloczkowski and Yaohang Li. Proceedings of BIBM 2021. Houston, TX, USA: IEEE, 2021, s. 3075-3082. ISBN 978-1-6654-0126-5. Dostupné z: https://dx.doi.org/10.1109/BIBM52615.2021.9669418.

Základní údaje

Originální název

An Innovative Perspective on Metabolomics Data Analysis in Biomedical Research Using Concept Drift Detection

Název česky

An Innovative Perspective on Metabolomics Data Analysis in Biomedical Research Using Concept Drift Detection

Autoři

SCHWARZEROVÁ, Jana (203 Česká republika), Adam BAJGER (203 Česká republika, domácí), I. PIERDOU, Lubomír POPELÍNSKÝ (203 Česká republika, domácí), Karel SEDLÁŘ (203 Česká republika) a W. WECKWERTH

Vydání

Houston, TX, USA, Proceedings of BIBM 2021, od s. 3075-3082, 8 s. 2021

Nakladatel

IEEE

---

Další údaje

Jazyk

angličtina

Typ výsledku

Stať ve sborníku

Obor

10602 Biology , Evolutionary biology

Stát vydavatele

Česká republika

Utajení

není předmětem státního či obchodního tajemství

Forma vydání

elektronická verze "online"

Kód RIV

RIV/00216224:14330/21:00126185

Organizační jednotka

Fakulta informatiky

ISBN

978-1-6654-0126-5

DOI

http://dx.doi.org/10.1109/BIBM52615.2021.9669418

Klíčová slova česky

Machine Learning; Concept drift; Metabolomics Analysis; Biomedical engineering; Computational biomedical analysis; Metabolomic prediction

Klíčová slova anglicky

Machine Learning; Concept drift; Metabolomics Analysis; Biomedical engineering; Computational biomedical analysis; Metabolomic prediction

Příznaky

Mezinárodní význam, Recenzováno

---

Anotace

V originále

The most challenging applications of data analysis prediction are mostly related to scenarios, where the source data is being provided in a time course. As the distribution of the underlying reality shifts over a time, a classification model trained on the previously relevant data starts to yield incorrect predictions about the data that are relevant right now. This phenomenon in machine learning is called concept drift. Within biomedical data, one of the molecular networks that is most significantly changing over a time, is the metabolome. Using metabolomics analysis to biomedical applications, makes an ideal tool for preventive healthcare, pharmaceutical industry, and even ecology engineering. This study provides an innovated perspective on the analysis of metabolomics datasets using the concept of drift detection. The evaluation is based on two main goals. The first goal is connected to the concept drift detection in available metabolomics datasets and the second goal is to provide the assessment of commonly used tools, resulting in the best detection approach for a general metabolomics dataset.

Česky

The most challenging applications of data analysis prediction are mostly related to scenarios, where the source data is being provided in a time course. As the distribution of the underlying reality shifts over a time, a classification model trained on the previously relevant data starts to yield incorrect predictions about the data that are relevant right now. This phenomenon in machine learning is called concept drift. Within biomedical data, one of the molecular networks that is most significantly changing over a time, is the metabolome. Using metabolomics analysis to biomedical applications, makes an ideal tool for preventive healthcare, pharmaceutical industry, and even ecology engineering. This study provides an innovated perspective on the analysis of metabolomics datasets using the concept of drift detection. The evaluation is based on two main goals. The first goal is connected to the concept drift detection in available metabolomics datasets and the second goal is to provide the assessment of commonly used tools, resulting in the best detection approach for a general metabolomics dataset.