@article{SORRIBES20241452,
title = {{ML-based} predictive gut microbiome analysis for health assessment},
journal = {Procedia Computer Science},
volume = {239},
pages = {1452-1459},
year = {2024},
note = {CENTERIS – International Conference on ENTERprise Information Systems / ProjMAN - International Conference on Project MANagement / HCist - International Conference on Health and Social Care Information Systems and Technologies 2023},
issn = {1877-0509},
doi = {https://doi.org/10.1016/j.procs.2024.06.318},
url = {https://www.sciencedirect.com/science/article/pii/S1877050924015618},
author = {Manel Gil Sorribes and Gabriele Leoni and Antonio Puertas Gallardo and Mauro Petrillo and Sergio Consoli and Vicenç Gómez and Mario Ceresa},
keywords = {GMHI, Gut Microbiome, Metagenomics, Machine Learning, Neural Network, COVID-19},
abstract = {Personalised medicine is a rapidly evolving field to which many resources have been devoted recently. It represents a paradigm shift from a one-size-fits-all approach to healthcare, focusing instead on tailoring treatments and diagnoses to individual patients. This study aims to contribute to this transition by leveraging recent advancements in microbiome research. An earlier study that computed the Gut Microbiome Health Index (GMHI), a potent indicator capable of predicting disease presence with approximately 70% of accuracy, serves as the foundation for this research. Positive values of the GMHI are associated with healthy subjects and negative values to non-healthy ones. The objective of this study is twofold: firstly, to advance the use of the GMHI through the application of two distinct machine learning techniques, a Fully Connected Neural Network and an Autoencoder, secondly, to adapt the GMHI for a unique dataset of COVID-19 patients. The employment of these two neural network architectures facilitated an enhancement in predictive accuracy to approximately 74.5%, surpassing the GMHI baseline accuracy of 70.95%. Simultaneously, the application of the GMHI, recalibrated using species specifically identified in the COVID-19 dataset, demonstrated a substantial increase in accuracy by 17% (achieving an accuracy of 76%). These promising results in distinguishing COVID-19 patients, highlight the potential and broad applicability of this approach in the sphere of personalized medicine and directly relating the COVID-19 disease to one’s microbiome and its composition.}
}