Forensic microtrace investigation relies on a time- and labour-intensive process of manually analysing samples via microscopy. To aid forensic experts in their investigations, an image recognition model for microtrace localisation and classification is needed. This work investigates the trace recognition accuracy that can be achieved by analysing images captured with automated microscopy through deep learning. Fibres, hairs, skin, glass and sand are pixel-wise classified in microscopy scans of tape-lift samples. As deep learning requires extensive amounts of annotated training data, pretraining is investigated to minimise the required annotation workload. ImageNet pretraining, pretraining with self-supervised learning and a sequential application of these approaches are tested. It is found that pretrained models are able to reduce the required annotated data twofold compared to models trained from scratch, while retaining the prediction accuracy. While the ImageNet-pretrained models outperform the self-supervised-pretrained models, the highest accuracy is achieved by combining the two approaches. With this, traces are recognised with a mean intersection over union of 0.56 when training on only 2.2 dm2 of annotated tape lift scans.

In animal cells, cell shape is primarily regulated by the actin cortex, a thin filament network connected to the plasma membrane. The architecture of the cortex is considered a key regulator of its function. Visualising the cortex is difficult due to the high density of numerous small-sized proteins involved in its formation. Consequently, the structure of the cortex at the membrane remains poorly studied. This study aims to gain insight into the organisation at the membrane of two key cortical components, human septin and septin-recruited actin. To study these filament structures, we reconstitute minimal cortices on supported lipid layers as model-membranes, allowing for imaging with electron and atomic force microscopy. We show that membrane binding of human septin results in ordered organisations of filaments. However, we find septin organises into arrays of paired filaments when incubated on a lipid monolayer and networks of bundles when incubated on a lipid bilayer. In addition, we showed a proof of concept for actin cortex reconstitution on lipid monolayers, which allowed us to see actin recruitment by septin meshworks.

Production and manipulation of microdroplets is an active are of research. The demand for finding new ways to produce microdroplets has resulted in a challenge to characterize these droplets. The problem becomes especially difficult for microdroplets which are placed on a transparent hydrophilic substrate. Ink-jet printing, Digital microfluidics, DNA synthesis are some of the applications which need topography measurements of low contact angle microdroplets. One application which is of interest is the sample preparation for Cryo Electron Microscopy (cryo-EM). A recent development in the search for efficient sample preparation for cryo-EM is to use hollow microcantilevers (HMC). HMCs can isolate even a single sub-cellular component and help prepare samples confined in a femtoliter droplet. These droplets are dispensed on a hydrophilic Electron Microscopy grid (EM-grid). By controlling the thickness of the water layer on the EM-grid through evaporation, it is possible to make the HMC technology more reliable by reducing sample wastage. However, such a control loop is absent. A real-time topography measurement of such droplets can serve as a control signal which can be used as feedback for a control loop. The objective of this project is to investigate the feasibility of optical methods to measure the topography information of a low contact angle microdroplet. The scope of this project is limited to develop a tool for a droplet which is supported by a glass slide (droplet-on-glass). In a subsequent study, the tool will be tested for droplets supported on grid (droplet-on-grid). An optical interferometry setup is proposed as a solution. A Mach-Zehnder interferometer is built to obtain the experimental fringes. A Single Frame Fourier Transform technique was used to analyze the data and obtain the results. Droplets of glycerol were dispensed on a glow discharged glass slide. Topography of the droplet was obtained until complete evaporation and a detection limit of 165nm was achieved. The accuracy of the method was found by comparing the results obtained with a Bruker White Light Interferometer (Bruker-WLI) in Phase Shifting Interferometry mode. An accuracy of 23\% was observed. The proposed setup had a repeatability of 14.7nm. To measure the reproducibility of the setup, a 3d printed structure which had the same size and shape as that of a droplet was used. The reproducibility of the setup was found to be 19.8nm over three days. Simulations were performed to analyze the effect of filter shape, filter width and the carrier frequency. Based on these findings, steps to measure a droplet-on-grid system with the proposed setup is explained. Further, to test the capabilities of the instrument the evaporation of a large water droplet on a EM-grid was observed using the proposed setup. It was found that the motion of fringe pattern as the droplet evaporates could give a good indication to control the evaporation time of even conventional machines, which do not employ hollow microcantilevers to dispense small droplets. However, final validation of the proposed setup with the cryo-EM is yet to be performed due to time constraints. As a recommendation for future work, new ways of dispensing samples on the EM-grid are explored. The necessary steps required to validate the setup with a cryo-EM are explained. Further, some ways in which the analysis time could be reduced are explored. This will be helpful in developing a software necessary to implement a real-time solution.

In eukaryotic cells, the segregation of nuclear information from the cytosol is carried out by the Nuclear Envelope, a double lipid membrane which protects the genes from cytoplasmic mutagens. To establish a connection between the cytoplasm and nucleoplasm, tiny pores named Nuclear Pore Complexes (NPCs) span through such membrane. The biological function of the NPC consists in regulating the bidirectional transportation of small molecules, like ions and transcriptional proteins, between the two compartments by imparting a selective barrier. FG-Nups are key proteins in this process, as they line the inner channel of the NPC and are responsible for the selective behaviour. Despite many models have attempted to explain the mechanism behind selectivity and substantially agree on the importance of FG-Nups in the context of nuclear transport, the exact details about how FG-Nups arrange within the NPC lumen and interact with translocating cargoes are still debated. In the current project, we illustrate an in-vitro approach to study the behaviour of purified FG- Nups, that takes advantage of DNA-origami nanotechnology. We show that it is possible to recreate a minimalistic version of the NPC by using an octagonally shaped DNA-origami as a scaffold while keeping low concentrations of magnesium in solution (< 1mM). The NPC mimic was built by coupling FG-Nups to specified locations along the inner wall of the origami. We characterized the DNA-origami structure, under different buffer conditions, before and after functionalization, by means of negative staining transmission electron microscopy and liquid atomic force microscopy. A volumetric comparison between the bare and FG Nup-coated DNA-origami provide quantitative confirmation of the success of the protein administration. We show that the incubation of a 10 times excess of proteins per binding site on the DNA-origami successfully couples the complexes together. Finally, we present our preliminary attempts at inserting a cholesterol-modified version of the DNA-origami molecule into both large and small vesicles, as well as planar lipid bilayers.