Human RPA (replication protein A), a single-stranded DNA-binding protein, is required for many cellular pathways including DNA repair, recombination and replication. However, the role of RPA in nucleotide excision repair remains elusive. In the present study, we have systematically examined the binding of RPA to a battery of well-defined ssDNA (single-stranded DNA) substrates using fluorescence spectroscopy. These substrates contain adducts of (6-4) photoproducts, N-acetyl-2-aminofluorene-, 1-amino-pyrene-, BPDE (benzo[a]pyrene diol epoxide)- and fluorescein that are different in many aspects such as molecular structure and size, DNA disruption mode (e.g. base stacking or non-stacking), as well as chemical properties. Our results showed that RPA has a lower binding affinity for damaged ssDNA than for non-damaged ssDNA and that the affinity of RPA for damaged ssDNA depends on the type of adduct. Interestingly, the bulkier lesions have a greater effect. With a fluorescent base-stacking bulky adduct, (+)-cis-anti-BPDE-dG, we demonstrated that, on binding of RPA. the fluorescence of BPDE-ssDNA was significantly enhanced by up to 8-9-fold. This indicated that the stacking between the BPDE adduct and its neighbouring ssDNA bases had been disrupted and there was a lack of substantial direct contacts between the protein residues and the lesion itself. For RPA interaction with short damaged ssDNA, we propose that, on RPA binding, the modified base of ssDNA is looped out from the surface of the protein, permitting proper contacts of RPA with the remaining unmodified bases.

<p>Recent advances in sensor technologies and data acquisition platforms have led to the era of Big Data. The rapid growth of artificial intelligence (AI), computing power and machine learning (ML) algorithms allow Big Data to be processed within affordable time constraints. This opens abundant opportunities to develop novel and efficient approaches to enhance the sustainability and resilience of Smart Cities. This work, by starting with a review of the state-of-the-art data fusion and ML techniques, focuses on the development of advanced solutions to structural health monitoring (SHM) and metamaterial design and discovery strategies. A deep convolutional neural network (CNN) based approach that is more robust against noisy data is proposed to perform structural response estimation and system identification. To efficiently detect surface defects using mobile devices with limited training data, an approach that incorporates network pruning into transfer learning is introduced for crack and corrosion detection. For metamaterial design, a reinforcement learning (RL) and a neural network based approach are proposed to reduce the computation efforts for the design of periodic and non-periodic metamaterials, respectively. Lastly, a physics-constrained deep auto-encoder (DAE) based approach is proposed to design the geometry of wave scatterers that satisfy user-defined downstream acoustic 2D wave fields. The robustness of the proposed approaches as well as their limitations are demonstrated and discussed through experimental data or/and numerical simulations. A roadmap for future works that may benefit the SHM and material design research communities is presented at the end of this dissertation.</p><br>