Valutazione di tecniche di individuazione e decodifica per l'identificazione robusta di regioni genomiche in nuvole di punti ottenute da Single-Molecule Localization Microscopy

Chromosome 3D organization is highly correlated with the nuclear behavior of the cell, for this reason there is a great interest in observing the former. Among the various techniques that are used, we focus on Single-Molecule Localization Microscopy. This method images a genomic region in such a way that the output representation is a cloud of points. Moreover when dealing with long sequences a hierarchical SMLM technique is used to speed up the process. To infer a genomic region's position one needs to identify and remove noise from the point-clouds and, in the case of the hierarchical technique, decode the noise-cleaned regions. To overcome the lack of labeled data we first developed an SMLM simulator. We then used generated and real data to evaluate first a set of clustering methods, some of which are novel in the field, on the removal of noise, and then a decoding technique and its variations on the identification of regions from hierarchical imaging.