Digital Bifurcation Analysis

Samuel Pastva

SAMUEL PASTVA

Faculty of Informatics, Masaryk University
  • Mgr. in Parallel and Distributed Systems, graduated with honours in 2017.
  • Member of the Sybila systems biology laboratory since 2015.
  • PhD student and researcher since 2017, with prof. Luboš Brim as advisor.

ACADEMIC BACKGROUND

SYSTEMS BIOLOGY

  • Biological systems can be studied via computational models.
  • Models are predictive and explainable.
  • In silico analysis is faster and cheaper than in vivo experiments.

BIFURCATION ANALYSIS

  • Models often contain unknown or uncertain elements ⇒ parameters.
  • Bifurcation analysis: inspect which parameters (and for what values) lead to significant change in model behaviour.

RESEARCH PROBLEM

  • Bifurcation is only defined in quantitative models (how much?), not qualitative (yes/no?) models.
  • For large systems, qualitative (or hybrid) approach is often necessary.
  • How to study bifurcation in qualitative models?

?

METHODOLOGY

  • Qualitative models correspond to directed graphs, or sets of runs over directed graphs.
  • Long-term behaviour of the graph corresponds to SCCs (strongly connected components), especially BSCCs.
  • Bifurcation analysis: identify which parameter values lead to a change in SCC/BSCC structure.

METHODOLOGY

  • Development of new symbolic algorithms that can detect (B)SCCs in very large graphs with parameters (up to 2^100 - 2^1000 states).
  • A new interactive visualisation technique based on decision trees for identification of critical parameter values.

Application

  • AEON: Bifurcation analysis of Boolean networks.
  • Online editor of parametrised models with static validation of structural properties.
  • Compute types of possible behaviour and explore dependence on parameters as a decision tree.

FUTURE PROSPECTS

  • Case study showcasing bifurcation analysis of interferon type-1 inflammation model.
  • New collaboration started with Faculty of Medicine at MUNI.
  • AEON used by students of bioinformatics at MUNI and VUT.
  • Possible future collaboration with Boolean network research at JAIST in Ishikawa.

other EXPERIENCE

  • Co-author of 15 ranked conference papers (7 A-ranked) and 4 journal papers.
  • Supervisor of three successful Bc. students, two of these later resulted in conference publications.
  • Artefact evaluation chair (CMSB '18,'21), program committee member (CIBCB, ICONS) and reviewer for other conferences and journals.
  • Seminar tutor for Computability and Complexity as well as Advanced Algorithms and Data Structures courses since 2015.
  • Oracle Labs research internship on VM tooling team (2020).
  • Brno PhD Talent scholarship (2018-2020).