Скачать или смотреть POWER OF PREDICTION OF ASTHMA USING GENETIC MARKERS OF THE RORA GENE

Título:
POWER OF PREDICTION OF ASTHMA USING GENETIC MARKERS OF THE RORA GENE

Autores:
Camila Alexandrina Viana de Figueiredo, LUCIANO GOMES

Descrição:
Asthma is a very heterogeneous chronic inflammatory airway disease. A gene associated with asthma and other allergy-related diseases has been excessively studied by groups of researchers: RORA (retinoic acid receptor-related orphan receptor alpha). Due to the complexity of asthma, advanced artificial intelligence studies are the key to try to decipher patterns for complex decision making. In this sense, machine learning algorithms are useful tools for predicting complex diseases. This study aims to scan some machine learning algorithms to predict asthma and non-asthma and discuss their particularities. This work is a pilot study in which we used genetic data from subsamples from a larger genetic bank, which comprises 535 genotyped SNPs (Illumina Multi-Ethic AMR/AFR Kit BeadChip) of individuals admitted to the ProAR. All data processing and results generation was performed by R. The tested algorithms were the Naive Bayes classifier and the decision tree: C5.0, 1R, and RIPPER. Naive Bayes algorithm presented low accuracy of its test prediction (58.50%) when we used all genectic markers. In this sense, the search for more informative markers for prediction is adequate. The C5.0 algorithm showed 97.8% training accuracy, however, its test accuracy drops to 60.89%. At this point, C5.0 signals the main markers used for decision-making. The 1R algorithm returns only one marker that it considers the best to classify the outcome. Of these results, three SNPs shared the findings by C5.0: rs1550225, rs140827468, and rs72750685. There were no similarities in decision-making between sequential 1R analisis and RIPPER. We returned to the analysis with Naive Bayes, however, using only the three SNPs taken together. We observed that the accuracy of the prediction was 73.13%. Algorithms work differently, but they can jointly signal genetic markers that can become the focus of other studies.

#asthma, #machine learning, #prediction, #SNP, #CongressoVirtualUFBA2021, #videoposter, #UFBA