Novel developments of atropisomeric organic molecules. Design of atropisomeric drugs, fluorescent compounds and investigation of reaction mechanisms.

In recent years, the study of restricted rotation bonds in organic compounds has aroused increasing interest. The reason is that this characteristic can lead to obtaining new properties in organic compounds. In this research thesis, an intense investigation was carried out using DFT calculations and experimental evaluation of the barriers to rotational energies, in order to discover new properties deriving from the restricted rotation bonds. Research has been developed in various fields of organic chemistry, ranging from drugs (the atropisomeric atorvastatin in Chapter 3) to luminescent compounds (aryls amino borane in Chapter 4). Furthermore, an organocatalytic central to axial conversion mechanism was investigated through DFT calculations, finding out interesting outcomes (Chapter 5). Finally, a project in collaboration with Dr. Farran and Prof. Vanthuyne of the Aix-Marseille University was done to investigate the interactions in transition states of rotational barriers.

Supervised and weakly supervised counting-by-segmentation: the fluorescent microscopy use case

This thesis focuses on automating the time-consuming task of manually counting activated neurons in fluorescent microscopy images, which is used to study the mechanisms underlying torpor. The traditional method of manual annotation can introduce bias and delay the outcome of experiments, so the author investigates a deep-learning-based procedure to automatize this task. The author explores two of the main convolutional-neural-network (CNNs) state-of-the-art architectures: UNet and ResUnet family model, and uses a counting-by-segmentation strategy to provide a justification of the objects considered during the counting process. The author also explores a weakly-supervised learning strategy that exploits only dot annotations. The author quantifies the advantages in terms of data reduction and counting performance boost obtainable with a transfer-learning approach and, specifically, a fine-tuning procedure. The author released the dataset used for the supervised use case and all the pre-training models, and designed a web application to share both the counting process pipeline developed in this work and the models pre-trained on the dataset analyzed in this work.