Plastic pollution has emerged as a global environmental challenge, prompting the need for innovative strategies to address the mounting accumulation of plastic waste, such as bioremediation through living organisms like bacteria, fungi, or plants to break down or neutralize pollutants in the environment. This thesis explores the promising avenue of plastic degradation through microbial action, focusing on the search for microbial enzymes capable of breaking down plastics, with a particular emphasis on polyethylene terephthalate (PET). The goal of this work is to develop machine learning models able to identify enzymes for PET degradation in a pool of available proteins. Protein sequence and structure serve as complementary sources of information for creating numerical representations for each protein under analysis. These numerical representations are then used to train semi-supervised classification models capable of distinguishing PET-degrading proteins from others. Experimental validations on a representative protein set yield high performances for all the tested models, particularly those that incorporate sequence information. The results suggest that these methods can detect crucial molecular markers associated with the ability to degrade PET in both information sources, allowing the prediction of unknown PET-degrading enzymes coming from microorganisms adapted in heavily plastic-polluted environments.
Plastic pollution has emerged as a global environmental challenge, prompting the need for innovative strategies to address the mounting accumulation of plastic waste, such as bioremediation through living organisms like bacteria, fungi, or plants to break down or neutralize pollutants in the environment. This thesis explores the promising avenue of plastic degradation through microbial action, focusing on the search for microbial enzymes capable of breaking down plastics, with a particular emphasis on polyethylene terephthalate (PET). The goal of this work is to develop machine learning models able to identify enzymes for PET degradation in a pool of available proteins. Protein sequence and structure serve as complementary sources of information for creating numerical representations for each protein under analysis. These numerical representations are then used to train semi-supervised classification models capable of distinguishing PET-degrading proteins from others. Experimental validations on a representative protein set yield high performances for all the tested models, particularly those that incorporate sequence information. The results suggest that these methods can detect crucial molecular markers associated with the ability to degrade PET in both information sources, allowing the prediction of unknown PET-degrading enzymes coming from microorganisms adapted in heavily plastic-polluted environments.
Machine learning models for the discovery of new plastics degradation enzymes
MINTO, STEFANO
2023/2024
Abstract
Plastic pollution has emerged as a global environmental challenge, prompting the need for innovative strategies to address the mounting accumulation of plastic waste, such as bioremediation through living organisms like bacteria, fungi, or plants to break down or neutralize pollutants in the environment. This thesis explores the promising avenue of plastic degradation through microbial action, focusing on the search for microbial enzymes capable of breaking down plastics, with a particular emphasis on polyethylene terephthalate (PET). The goal of this work is to develop machine learning models able to identify enzymes for PET degradation in a pool of available proteins. Protein sequence and structure serve as complementary sources of information for creating numerical representations for each protein under analysis. These numerical representations are then used to train semi-supervised classification models capable of distinguishing PET-degrading proteins from others. Experimental validations on a representative protein set yield high performances for all the tested models, particularly those that incorporate sequence information. The results suggest that these methods can detect crucial molecular markers associated with the ability to degrade PET in both information sources, allowing the prediction of unknown PET-degrading enzymes coming from microorganisms adapted in heavily plastic-polluted environments.File | Dimensione | Formato | |
---|---|---|---|
Stefano_Minto_Master_thesis (1).pdf
accesso aperto
Dimensione
2.01 MB
Formato
Adobe PDF
|
2.01 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/64794