Seminar

Our seminar

Laboratory seminars (PV273 in the course catalog) on Wednesday, 15:00-16:00, A505, FI MU, Botanická 68

The format of standard lectures: 30 minutes presentation + 15 minutes for questions, slides in English, presentation in English or Czech based on audience

Programme for Autumn 2018

  • 19.9.2018
    Ing. Jana Hozzová, Ph.D.
    Improving search heuristics in molecular docking
    Abstract: Before the drug binds into the active site of the protein, it may need to pass through the tunnel within the protein. Molecular docking, a computer simulation, looks for a specific position of the drug so that it fits into the tunnel at the current spot. In collaboration with Loschmidt Laboratories in Faculty of Science at Masaryk University, we have developed a software tool CaverDock which applies a modified docking to construct a complete, smooth trajectory of the drug passing through the tunnel. From the point of computer science, the process of construction of the trajectory includes searching in a vast space of all possible conformations of the molecule. In this talk, we explain the current heuristics of that search and our attempts to improve it.
  • 26.9.2018
    Cancelled due to MetaCenter seminar
  • 3.10.2018
    Mgr. Roman Lacko
    GitLab at Faculty of Informatics
    Abstract: Git is currently one of the most used version control systems, mainly due to its flexibility, abundance of features and support for distributed workflows. Repository managers, with Git at their core, aim to further improve software development workflows and help reduce the management overhead. Computer Systems Unit at Faculty of Informatics hosts one of such managers, GitLab. Since its deployment in April 2016 several programming courses and laboratories have migrated their materials there. It currently hosts over 5000 projects and 1700 users. CSU FI is searching for ways to help the staff to integrate the ever increasing number of features to their courses and research.
    PDF
  • 10.10.2018
    Mouzhi Ge, Ph.D.
    Enhance protein chain similarity searching performance through data science techniques
    Abstract: Searching similar protein chains is very helpful for the research of biologists. However, due to the exponentially increasing amount of protein structural data and the complexity of the 3D protein structure. It is very expensive to perform the protein chain comparison by traversing the whole Protein Data Bank (PDB). Therefore, we intend to solve this performance problem via data science techniques such as dimension reduction, data clustering, data classification as well as new algorithms based on pre-computing results. In this talk, I will first clarify the research problem and define the roadmap of this project from the data science perspective. Then the key part of this talk is about how to approach this problem and how to use the similar methods to solve other problems. As this is a research-in-progress work, I will report the state-of-the-art results we have achieved. Also, this talk will touch base on planned research in the future and ideas for further discussions.
  • 17.10.2018
    RNDr. Milan Čermák
    Stream4Flow: Stream-based IP flow analysis using Apache Spark
    Abstract: IP flow analysis is an essential part of traffic measurement for cyber security, especially in high-speed, large-scale networks. Although the widespread use of this approach, we still face to issues associated with increasing volume of IP flows, insufficient scalability of current data analysis systems, and demands to minimize the time required for cyber threats detection. In this talk, I will provide an overview of current IP flow analysis systems and introduce our stream-based approach that overcomes the issues. The approach will be demonstrated on the Stream4Flow analysis framework utilizing Apache Spark engine and other modern systems for big data processing. I will show that the framework not only allows to implement common detection methods as scalable real-time applications but also provides new insight to IP flow data.
  • 24.10.2018
    RNDr. Miloš Liška, Ph.D.
    How we are transmitting live uncompressed 8K video and still not building corresponding SDN networks
    Abstract: After a successful demonstration of live, JPEG compressed 8K video transmission implemented using the UltraGrid software over commodity academic networks, we continue with the development of the UltraGrid software towards supporting transmissions of an uncompressed 8K video. Uncompressed 8K video with 60 frames per second introduces a bitrate of approximately 33 Gbps which is unmatched for any other video streaming. In the talk, we provide the current state of our implementation of uncompressed 8K video transmission in UltraGrid including hardware and software considerations of such implementation. In the second part of the talk, we discuss the current state of reaching for end-to-end connectivity providing bandwidth above 10 Gbps within the CESNET network and current issues of reaching out with uncompressed 8K video transmissions internationally. We are convinced that a possible solution would be a deployment of Software Defined Networking allowing reservation of end-to-end circuits with defined bandwidth over 40Gbps and high-speed network links.
    PDF
  • 31.10.2018
    Mgr. Zdenka Dudová, Ph.D.
    Building an IT infrastructure for a specific research community
    Abstract: In cooperation with EurOPDX consortium focused on Patient-derived tumor xenograft (PDX) models, we got an H2020 project focused on building a research infrastructure (RI) for starting communities. Surprisingly our project partners who are doing top-level research use excel sheets and papery documentation during their work. Sending sensitive data by emails is not unusual too. This talk should be about implementation of information technologies into well-established workflows at all levels of PDX research concerning data mining, harmonization, storage, processing, analysis, dissemination, etc. The talk will introduce RI design and show a possible way how to uplift the conditions for PDX community.
    The value added of this effort will be the automatization. The data from Laboratory Assistant Suits (LAS, developed by University of Torino) will be automatically uploaded to a data repository where will be stored other consortial data in several formats available for all tools (e.g., cBioPortal, GenomeCruzer, PDX Finder) integrated into the RI. For users there will be an interface enabling browsing the consortial data, filtering and visualizing them by the integrated tools and after will be able to apply for samples which arose from their search. Data upload should be automatized the same as whole data flow through the RI including data harmonization. The whole infrastructure will be covered by AAI to protect unpublic data and get insight into the backgrounds of the infrastructure users.
  • 7.11.2018
    RNDr. Dalibor Klusáček, Ph.D.
    Batch job scheduling using automated job runtime predictions
    Abstract: While the main concepts of batch job scheduling are well known and used for decades, some crucial problems remain aside although their importance is obvious. One of such problems is the known inaccuracy of user-provided job runtime estimates. These estimates are very critical during many phases of job scheduling process. In this talk, we demonstrate the importance and impact of (in)accurate job runtime estimates and present our recent research results focusing on automated runtime-prediction techniques. This presentation is a joint work between CESNET (MetaCentrum) and the Karlsruhe Institute of Technology (Steinbuch Centre for Computing).
  • 14.11.2018
    Ing. Martin Žádník, Ph.D.
    DDoS protection – mitigation algorithms
    Abstract: The volume of Distributed Denial of Service (DDoS) attacks and their variety grows every year. Since 2016 the largest attacks reached 1 Tbps, effectively disconnecting even well provisioned services from the Internet. CESNET decided to exploit its expertise in building hardware-accelerated network probes to build its own active device with mitigation capabilities. However, attackers often adapt to the protection introduced and therefore various mitigation algorithms must be set in place to be applied according to the attack type. The talk will shortly introduce the CESNET solution and subsequently will focus on the anatomy of attacks and their mitigation algorithms. The presentation will finish with an outline of several challenges in the mitigation domain.
  • 21.11.2018
    doc. RNDr. Barbora Kozlíková, Ph.D.
    Visualization and Visual Analysis of Proteins and their Dynamics
    Abstract: Understanding of protein structure, function, and behavior is crucial for the biochemists to design new chemical compounds, forming, for example, a basis of a new medication. In our long-term collaborative research with the protein engineers, we are aiming to provide them with the visual guidance through their complex data and tasks. One of the most challenging tasks is to explore the large simulations of molecular dynamics, which can consist of hundreds of thousands of time steps. In this talk, I will present several examples of how to support the exploration process by several visualization techniques, designed specifically for this purpose.
    PPTX
  • 28.11.2018
    RNDr. Tomáš Rebok, Ph.D.
    Project ANALYZA — Distributed Framework for Large-scale Data Analysis
    Abstract: Our talk will present the distributed framework for large-scale data analysis, being developed within the ANALYZA project. The framework — consisting of a large-scale data store, distributed processing modules and various data visualizations — is mainly intended for police investigation purposes, thus its selected proof-of-concept use-cases will be presented as well. Finally, we present an idea about scalable adaptive framework for general data analysis purposes, which is being researched by the BigData group at the Institute of Computer Science MU.
  • 5.12.2018
    Mgr. Michal Pajtinka
    Detection of geometrical arrangements of atoms of molecular fragments
    Abstract: Databases storing 3D structures of molecules are an important source of data used by life sciences — impressive scientific outputs have been published and several Nobel prizes awarded for research based on this data. However, the quality and reliability of these data are crucial for research. Protein Data Bank (PDB), the most well-known database for structural data of biomacromolecules, has previously integrated deposition system OneDep that also performs input data validation. Validation reports generated by that process provide us with useful information about missing atoms, atom bond length outliers, unexpected bond angles, wrong chirality, and similar issues. Although these reports have already significantly improved the quality scores for proteins in PDB database, the system is still being under active development, and some issues are not being checked at all — among other things, shapes of some molecular fragments (rings, scaffolds…) in protein ligands. Previous research shows that some rings have their atoms geometrically arranged in an energetically unfavorable manner in more than 20% of their occurrences. In cooperation with PDBe (European branch of Worldwide Protein Data Bank organization), we are developing a solution to detect this problem.
  • 12.12.2018
    Mgr. Lukáš Ručka
    Adopting Method of Manufactured Solutions for real-time biomedical simulations (rehearsal for thesis proposal defense)
    Abstract: The Method of Manufactured Solutions is a well established tool in the area of verification of computational fluid dynamic codes. However, using the method on state-of-the-art real-time soft tissue simulators faces challenges, both theoretical and practical. Furthermore, the method requires the use of apriori chosen analytic solutions, which are rarely available. The talk addresses the challenges and why it is actually a good idea to use solutions broken by design (and what they offer for patient-specific modeling).

    and Christmas meeting

  • Examination period 2.1. – 15.2.2019
    Meetings devoted to state exam training are scheduled mostly. State exam training are in Czech mostly. More days can be added if needed by students.

  • 16.1.2019
    State exam training

    TBA
  • 23.1.2019
    State exam training

    Martin Brakl: Srovnání dostupných řídících kontrolérů softwarově-definovaných sítí
    Vedoucí: Tomáš Rebok
    Oponent: Miloš Liška
  • 30.1.2019
    State exam training

    Natália Gregušková: Kryptoadresový server
    Vedoucí: Michal Zima
    Oponent: David Střelák

    Zoltán Fridrich: Plugin pro přehrávání videa pro framework SOFA
    Vedoucí: Lukáš Ručka
    Oponent: Miloš Liška

Past seminars

Contact: Hana Rudová

(c) 2018 SITOLA, Administration