Banca de QUALIFICAÇÃO: MARCELA DE ANGELIS VIGAS PEREIRA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
STUDENT : MARCELA DE ANGELIS VIGAS PEREIRA
DATE: 15/07/2024
TIME: 08:00
LOCAL: https://meet.google.com/drh-dwvx-fji
TITLE:

USE OF MACHINE LEARNING IN THE IDENTIFICATION OF TAXONOMIC PATTERNS ASSOCIATED WITH THE GUT MICROBIOME IN PARKINSON’S DISEASE

KEY WORDS:

Machine Learning; Classifier Committee; Parkinson's Disease; Gut Microbiome; Metagenomics

PAGES: 40
BIG AREA: Ciências Biológicas
AREA: Biologia Geral
SUMMARY:

Parkinson's disease is a pathological and progressive condition predominantly affecting the elderly. It is primarily characterized by tremors, which were initially considered the primary manifestation of the disease. Currently, it is known that the effects of this disease are not limited to voluntary movements, as gastrointestinal changes, such as constipation, have been observed decades before the onset of tremors. This process originates from alterations in the gut microbiota of individuals affected by Parkinson's disease, which has sparked scientific interest in better understanding Parkinson's disease and developing new diagnostic and intervention proposals. Among these areas of investigation are metagenomics and machine learning, which together can contribute to the discovery of patterns in the composition of the gut microbiota associated with Parkinson's disease. Thus, the present study aims to analyze taxonomic patterns related to the disease through a machine-learning model characterized as a homogeneous committee of Random Forest classifiers trained with metagenomic data associated with Parkinson's disease. So far, the proposed model has shown satisfactory performance, obtaining performance metrics (AUC of 82%, execution time of 5.40 seconds, and F1-score of 69%) that resemble or surpass those of some previous studies conducted with a single Random Forest classifier. It was also able to list as relevant attributes genera such as Roseburia and Clostridium, cited as important in previously found taxonomic patterns.

COMMITTEE MEMBERS:
Presidente - 3083298 - RENAN CIPRIANO MOIOLI
Interno - 1513597 - JOAO PAULO MATOS SANTOS LIMA
Externo ao Programa - 3884005 - PATRICK CESAR ALVES TERREMATTE - UFRN