Studentská vědecká konference

The development of therapeutic monoclonal antibodies (mAbs) is a multi-billion-dollar industry with nearly a hundred therapeutic antibodies currently approved for use in the EU and much more waiting for approval. However, the journey of a therapeutic mAb from discovery to the market is long and costly and can be marred by unsuitable properties of the antibody, such as aggregation or low stability. Developability index is a metric that aims to predict the antibody’s likelihood of reaching clinical stages. It can be calculated from the three-dimensional structure of the antibody, but the structure is often not available and its modeling can be computationally demanding and imprecise. Therefore, we aim to predict the developability index of antibodies directly from the amino acid sequence using machine learning. Various preprocessing and feature extraction methods are applied to a set of antibody sequences with known structures in order to obtain multiple training datasets. Different machine learning models are then compared in terms of classification performance.

Mycobacterium tuberculosis is a pathogenic bacterium and can cause tuberculosis. Because this bacterium has a poorly permeable cell wall, which complicates the treatment of this disease, it is essential to focus on GT-A glycosyltransferases, which this bacterium uses, among other things, to build its cell wall. The reasons mentioned above make them an interesting therapeutic target for fighting this bacterium. It is important to identify yet undescribed glycosyltransferases in the genome of M. tuberculosis CDC1551. For their retrieval, the program Phyre2 is used, which is a program that deals with the prediction of the 3D structure of proteins from a sequence. This program also has a feature that allows you to use a protein with an already solved 3D structure to scan a particular genome and in this way find proteins that are likely to be structurally similar. From the proteins found, some hypothetical proteins were finally selected and further investigated using the functions of the Phyre2, BLAST, and RaptorX programs to predict their likely function and to determine whether they could be GT-A glycosyltransferases.

Autoimmune diseases are pathologic conditions in which the reactivity of the immune system occurs in the form of autoantibodies and T-cell responses to self-structures. Autoimmune conditions, depending on the localization and severity of the process, lead to various clinical manifestations up to disability and death. To prevent the respond of the T-cells on the tissues of their own body, maturing T-cells are checked in the thymic medulla for interaction with the vast majority of self-antigens, and in the case of strong interaction, are subjected to negative selection. In some autoimmune diseases, especially in such a severe one as multiple sclerosis, there is an abnormal activation of the expression of human endogenous retroviruses (HERVs). HERVs are sequences of viral origin, which compose about 8% of our genome and are traces of infections that affected the primates' germ line along the last 100 million of years. In this work, we process bulk and single cell RNA-seq of mouse and human thymus cells to analyze autoantigens of retroviral origin, which are presented to maturing T-cells during negative selection.  

Genetic information in organisms is stored in a form of a DNA molecule. The information in the genome is not random. Therefore, certain patterns can be distinguished using various techniques. Gene coding and non-coding region determination from nucleotide sequence is the main focus of this work. Tokenization, a data compression algorithm, seem to be a useful tool in the process of understanding the language of genomes.

Structural and mechanical properties play a significant role in biological function of nucleic acids. Sequence-dependent structure of nucleic acids determines ligand binding, nucleosome positioning, promotor location, gene regulation and chromatin organisation. In this contribution, the conformation of nucleic acids is considered at the level of the rigid base approximation. In the rigid base model, nucleic acids are described using internal coordinates defining the relative displacement and orientation of bases within and between base pairs. Programs in C and Python were developed for efficient computation of these coordinates from molecular dynamics simulation data. The accuracy of the coordinate computation was verified, and runtime was compared to the standard software packages. The C program is roughly ten times faster than the standard implementation.

This work focuses on structural properties of methylated and unmethylated DNA depending on the context of bases around the CpG dinucleotides. Methylation plays a key role in many biological processes, such as gene expression, immune responses and organism development and aging. A suitable DNA sequence was divided into three parts. A copy of this sequence was then methylated. This way, six shorter sequences suitable for molecular dynamics simulations using AMBER software were obtained. Resulting trajectories were analysed by 3DNA software package. Data about local base-step parameters (shift, slide, rise, tilt, roll and twist) obtained by this analysis were processed in Python. Results suggest that there are changes to the local structure for distinct contexts, however, it is not possible to entirely rule out the impact of context context dependent population of BI/BII backbone substates.