Studentská vědecká konference

Melanoma phenotype switching is a process in which melanocytic cells are transformed into mesenchymal-like cell type.  Existing evidence suggests that this process is mediated by SOX10 decrease which is likely to prelude the acquisition of tumor's therapy-resistance. Aim of this thesis is to study chromatin accessibility and transcriptomic responses in individuals cells to transcription factor (TF) perturbations. It is achieved by utilizing modified CROP-seq (Datlinger et al., 2017) with multiomic readout on MM087 and MCF7 cell lines. SOX10, MITF and TFAP2A were selected as target TFs for the CRISPR-dCas9 system due to their role in melanoma phenotype switching. It is hereby shown that we were able to separate the two cell lines based on heterogeneity of their transcriptomic and epigenomic profile. Obtained results also suggest that MM087 cells with dCas9-MITF-KRAB gRNAs show significant SOX10 downregulation as the effect of MITF knock-down. Taken together, these analyses reveal gene expression and chromatin accessibility responses to regulatory perturbations in cancer cells and provide useful information for future gene regulatory networks reconstruction.

In the laboratory synthesis of a given compound, it is necessary to identify all available starting compounds and possible reaction pathways. The number of synthesis steps is an important parameter in drug design. It tells about the availability of the given substance, and it can also be used as a molecular complexity measure. This diploma thesis will deal with the prediction of the length of the synthetic pathway using machine learning methods. The aim of the work is to create suitable balanced data and with their help to implement a machine learning model to predict of the length of the synthetic pathway.

Mathematical models play an important role in predicting the progression and dynamics of disease spreading. Those predictions may serve to support policy decisions such as designing restrictions to stop or limit the spread. In recent years, mathematical models were commonly used in the fight against the pandemic of the SARS-CoV-2. Here, we limit our discussion to compartmentalised models that track the time evolution of several compartments (e.g. infected and susceptible) of the population. In this project, the aim is to explore the dynamics of the spread using several standard models. We study the behaviour of those models over a range of parameter values and assess how these parameters are linked to restrictions that can be put in place. The analysis proves that blocking the transmission by social distancing will have a significant effect on the rate of propagation through the population. Compartmentalised models, however, do not consider the spatial aspect of spreading. Therefore,  the models are extended to include mobility as a diffusive process. From those, it can be concluded that decreasing the rate of movement of infected individuals will reduce the speed of propagation. The analysis is performed using standard classical techniques and in silico using Python code.  

Human cancer is a heterogeneous disease initiated by random somatic mutations and driven by multiple genomic alterations. Cancer subtype classification has the potential to significantly improve disease prognosis and move closer toward personalized cancer treatment and prevention.  The assumption behind molecular subtyping is that patients of similar gene expression patterns are likely to have similar responses to therapies and clinical outcomes due to the tumor microenvironment. Thus, molecular subtyping can reveal information valuable for a range of cancer studies from cancer initiation and tumor biology to prognosis and personalized medicine. Molecular subtyping was widely described in breast and colorectal cancer, however, it was not broadly explored in other indications. In this work, molecular subtypes are explored via deep learning models, specifically variational autoencoder latent space. The analysis is performed on a range of publicly available datasets such as Gene Expression Omnibus (GEO), The Cancer Genome Atlas (TCGA) and the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC).

As a highly significant biomacromolecule, hyaluronic acid is present in all animal tissues as well as in body fluids. Its versatility and uniqueness are increasingly appreciated, especially in the field of medicine, while highly effective water binding is the key property. A detailed description of the dynamic behavior of HA surrounded by water molecules should make a great contribution to the development of already-established applications, as well as create inspiration for entirely new ways of use. Therefore, the Amber software package was utilized in conjunction with the GLYCAM06 force field to carry out molecular dynamics simulations of the disaccharide and tetrasaccharide building blocks of hyaluronan in water. Trajectories were subsequently analysed to describe the conformational dynamics of individual structures. Glycosidic bonds dihedral angles, ring puckering conformations, torsions of exocyclic groups, and emerging intermolecular and intramolecular hydrogen bonds were all investigated.

Recently many avian genes that were believed to be missing from avian genomes were found. These "missing" genes were mainly found in regions with high GC%. The reason a gene can be "hidden" can be twofold. It can either be truly absent (a mutation caused this gene to become a pseudogene), or the GC content causes technical obstacles to finding this gene at the sequencing level (polymerase slippage) or at the assembly level (overlapping long repetitive stretches). A similar problem could be in turtle genomes, which also have higher GC% compared to other vertebrates. The main goal of my work is to explore whether there are also hidden genes in turtle genomes. The way I approach this problem is through synteny analysis - blocks of co-occuring loci. Consider a block of genes labeled A, B and C. Gene B is absent from a turtle genome, but but genes A and C are present. Genes A and C have orthologs in other vertebrate taxa. We can compare the sequence between A and C with the sequences of B orthologs from other vertebraes. Either we can find a match - the gene is present but not yet annotated, or we can find a partial match with a gap - the gene is present but there is an assembly problem, or we end up not finding a match and have to try to find the gene sequence in the raw reads.

DNA methylation plays a key role in many biological processes, such as gene expression, immune responses, organism development, and aging. It mostly occurs in CpG dinucleotide steps where cytosine is replaced by 5-methylcytosine. Despite intense research, the effect of methylation on DNA structure and mechanical deformability is poorly understood. This work focuses on structural properties of methylated and unmethylated DNA depending on the context of bases around the CpG dinucleotides. Ten short sequences containing CpG in all tetranucleotide contexts, five methylated and five unmethylated controls, have been designed and their molecular dynamics (MD) simulated using AMBER software. So far, some of the MD trajectories have been processed by 3DNA software package to obtain time series of local base pair-step parameters (shift, slide, rise, tilt, roll and twist), which were further analysed. Results suggest that there are changes to the local structure due to distinct context-dependent populations of BI/BII backbone substates. However, it is not possible to entirely rule out the impact of sequence context on the conformation of the individual substates, and a deeper analysis is to be conducted.