Machine Learning (ML) is a branch of Artificial Intelligence (AI) that allow computers to learn without being explicitly programmed. Various are the applications of ML in pharmaceutical sciences, especially for the prediction of chemical bioactivity and physical properties, becoming an integral component of the drug discovery process. ML is characterized by three learning paradigms that differ in the type of task or problem that an algorithm is intended to solve: supervised, unsupervised, and reinforcement learning. In chapter 2, supervised learning methods were applied to extracts of Lycium barbarum L. fruits for the development of a QSPR model to predict zeaxanthin and carotenoids content based on routinely colorimetric analyses performed on homogenized samples, developing a useful tool that could be used in the food industry. In chapters 3 and 4, ML was applied to the chemical composition of essential oils and correlated to the experimentally determined associated biofilm modulation influence that was either positive or negative. In these two studies, it was demonstrated that biofilm growth is influenced by the presence of essential oils extracted from different plants harvested in different seasons. ML classification techniques were used to develop a Quantitative Activity-Composition Relationship (QCAR) to discover the chemical components mainly responsible for the anti-biofilm activity. The derived models demonstrated that machine learning is a valuable tool to investigate complex chemical mixtures, enabling scientists to understand each component's contribution to the activity. Therefore, these classification models can describe and predict the activity of chemical mixtures and guide the composition of artificial essential oils with desired biological activity. In chapter 5, unsupervised learning models were developed and applied to clinical strains of bacteria that cause cystic fibrosis. The most severe infections reoccurring in cystic fibrosis are due to S. aureus and P. aeruginosa. Intensive use of antimicrobial drugs to fight lung infections leads to the development of antibiotic-resistant bacterial strains. New antimicrobial compounds should be identified to overcome antibiotic resistance in patients. Sixty-one essential oils were studied against a panel of 40 clinical strains of S. aureus and P. aeruginosa isolated from cystic fibrosis patients, and unsupervised machine learning algorithms were applied to pick-up a small number of representative strains (clusters of strains) among the panel of 40. Thus, rapidly identifying three essential oils that strongly inhibit antibiotic-resistant bacterial growth.

Machine learning applications to essential oils and natural extracts

PATSILINAKOS, ALEXANDROS
2020

Abstract

Machine Learning (ML) is a branch of Artificial Intelligence (AI) that allow computers to learn without being explicitly programmed. Various are the applications of ML in pharmaceutical sciences, especially for the prediction of chemical bioactivity and physical properties, becoming an integral component of the drug discovery process. ML is characterized by three learning paradigms that differ in the type of task or problem that an algorithm is intended to solve: supervised, unsupervised, and reinforcement learning. In chapter 2, supervised learning methods were applied to extracts of Lycium barbarum L. fruits for the development of a QSPR model to predict zeaxanthin and carotenoids content based on routinely colorimetric analyses performed on homogenized samples, developing a useful tool that could be used in the food industry. In chapters 3 and 4, ML was applied to the chemical composition of essential oils and correlated to the experimentally determined associated biofilm modulation influence that was either positive or negative. In these two studies, it was demonstrated that biofilm growth is influenced by the presence of essential oils extracted from different plants harvested in different seasons. ML classification techniques were used to develop a Quantitative Activity-Composition Relationship (QCAR) to discover the chemical components mainly responsible for the anti-biofilm activity. The derived models demonstrated that machine learning is a valuable tool to investigate complex chemical mixtures, enabling scientists to understand each component's contribution to the activity. Therefore, these classification models can describe and predict the activity of chemical mixtures and guide the composition of artificial essential oils with desired biological activity. In chapter 5, unsupervised learning models were developed and applied to clinical strains of bacteria that cause cystic fibrosis. The most severe infections reoccurring in cystic fibrosis are due to S. aureus and P. aeruginosa. Intensive use of antimicrobial drugs to fight lung infections leads to the development of antibiotic-resistant bacterial strains. New antimicrobial compounds should be identified to overcome antibiotic resistance in patients. Sixty-one essential oils were studied against a panel of 40 clinical strains of S. aureus and P. aeruginosa isolated from cystic fibrosis patients, and unsupervised machine learning algorithms were applied to pick-up a small number of representative strains (clusters of strains) among the panel of 40. Thus, rapidly identifying three essential oils that strongly inhibit antibiotic-resistant bacterial growth.
17-dic-2020
Inglese
Machine learning; essential oils; natural extracts; qsar; qspr; qcar; cheminformatics; biofilm
RAGNO, Rino
VILLANI, Claudio
Università degli Studi di Roma "La Sapienza"
File in questo prodotto:
File Dimensione Formato  
Tesi_dottorato_Patsilinakos.pdf

accesso aperto

Dimensione 4.43 MB
Formato Adobe PDF
4.43 MB Adobe PDF Visualizza/Apri

I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.14242/88052
Il codice NBN di questa tesi è URN:NBN:IT:UNIROMA1-88052