Development of a Framework for the Classification of Antibiotics Adjuvants

Throughout the last decades, bacteria have become increasingly resistant to available antibiotics, leading to a growing need for new antibiotics and new drug development methodologies. In the last 40 years, there are no records of the development of new antibiotics, which has begun to shorten possib...

Full description

Saved in:

Bibliographic Details
Main Author:	da Silva, Carla Rafaela Mendes
Format:	Dissertation
Language:	English
Published:	ProQuest Dissertations & Theses 01-01-2022
Subjects:	Amino acids Antibiotics Antimicrobial agents Artificial intelligence Bacteria Bioengineering Biofilms Bioinformatics Biology Computer science Data base management systems Data mining Deoxyribonucleic acid DNA Drug resistance Extensible Markup Language Genetic engineering Genetics Information science Internet and social media studies Management Microbiology Molecular biology Penicillin Pharmaceutical sciences Pharmacology Ribonucleic acid RNA Software User interface Web studies World Wide Web
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Throughout the last decades, bacteria have become increasingly resistant to available antibiotics, leading to a growing need for new antibiotics and new drug development methodologies. In the last 40 years, there are no records of the development of new antibiotics, which has begun to shorten possible alternatives. Therefore, finding new antibiotics and bringing them to market is increasingly challenging. One approach is finding compounds that restore or leverage the activity of existing antibiotics against biofilm bacteria. As the information in this field is very limited and there is no database regarding this theme, machine learning models were used to predict the relevance of the documents regarding adjuvants.In this project, the BIOFILMad- Catalog of antimicrobial adjuvants to tackle biofilmsapplication was developed to help researchers save time in their daily research. This application was constructed using Django and Django REST Framework for the backend and React for the frontend.As for the backend, a database needed to be constructed since no database entirely focuses on this topic. For that, a machine learning model was trained to help us classify articles. Three different algorithms were used, Support-Vector Machine (SVM), Random Forest (RF), and Logistic Regression (LR), combined with a different number of features used, more precisely, 945 and 1890. When analyzing all metrics, model LR-1 performed the best for classifying relevant documents with an accuracy score of 0.8461, a recall score of 0.6170, an f1-score of 0.6904, and a precision score of 0.7837. This model is the best at correctly predicting the relevant documents, as proven by the higher recall score compared to the other models. With this model, our database was populated with relevant information.Our backend has a unique feature, the aggregation feature constructed with Named Entity Recognition (NER). The goal is to identify specific entity types, in our case, it identifies CHEMICAL and DISEASE. An association between these entities was made, thus delivering the user the respective associations, saving researchers time. For example, a researcher can see with which compounds"pseudomonas aeruginosa"has already been tested thanks to this aggregation feature.The frontend was implemented so the user could access this aggregation feature, see the articles present in the database, use the machine learning models to classify new documents, and insert them in the database if they are relevant.
ISBN:	9798381263299