Deep Learning for DNA Analysis
keywords ANTIBIOTICS, DRUG RESISTANCE, BACTERIA, DRUG DESIG, DEEP NEURAL NETWORKS, MACHINE LEARNING
Reference persons GIOVANNI SQUILLERO
External reference persons Alberto Tonda (INRAE, France); Pietro Barbiero (University of Cambridge, UK).
Research Groups DAUIN - GR-05 - ELECTRONIC CAD & RELIABILITY GROUP - CAD
Thesis type RESEARCH / EXPERIMENTAL
Description Convolutional neural networks (CNN), techniques from the domain of Deep Learning, are able to take into account relationships in sequences of data. Traditionally used for analysis of images and text, CNNs have been recently applied to the bioinformatics domain: for example, strands of RNA from SARS-CoV-2 (the virus responsible for COVID-19) have been scanned by a CNN to
uncover sequences that ultimately led to a reliable test.
CNNs could potentially be precious to tackle other medical issues, for example predicting the sensitivity to antibiotics for bacteria responsible for sepsis, an illness that is estimated to take over 11 million lives every year.
The objective of this thesis is to use a CNN to predict sensitivity or resistance to antibiotics for bacteria, analyzing their RNA as a sequence of data. Starting from real-world data, already collected from open repositories, the candidate shall prepare and test a CNN to verify whether such a machine learning system is able to predict sensitivity or resistance to antibiotics. In a second step, Explainable AI techniques will be used in an attempt to extract DNA sequences that could be validated by biologists to explain antibiotic resistance.
Deadline 31/07/2021 PROPONI LA TUA CANDIDATURA