National and Kapodistrian University of Athens
Development of computational methods to investigate the stochastic dynamic behavior of microbial communities
Abstract
dc:descriptionMethods and software tools for the analysis of cell movies which allow the accurate estimation of bacterial attributes at the single-cell level, the visualization and statistical modeling of the extracted cell attributes, and the creation of in silico experiments simulating microbial communities growth, remain currently a challenge in computational systems biology. Methods reported in the literature are lacking in many respects. They exhibit low success rates in cell segmentation of dense cell movies with thousands of cells, often require laborious parameterization, and are characterized as not friendly by user-biologists with no bioimage analysis expertise. Moreover, there is no end-to-end computational pipeline to automatically analyze cell movies, model the variability of extracted cell attributes, explore cell diversity visually, and facilitate the realistic simulation of microbial communities' behavior, while taking into account the inherent stochasticity of the underlying biological phenomena. This doctoral dissertation's main goal was to fill this gap by developing an end-to-end computational strategy with three stages for studying microbial communities' dynamic growth behavior. The first stage provides an automated bioimage analysis platform, which extracts, collects, and organizes the estimated cell attributes data hierarchically, without the need for human-user intervention. The second stage provides an information system for single-cell analytics and visualization. It allows the user to visualize in different ways and infer models for the cell attributes extracted by the image analysis stage at different community organization levels. The third stage supports the development of multiscale "digital twin" model realizations to investigate in silico microbial communities' growth. It can consider the individual cells' genetic "program", the microenvironment's conditions, and the inherent stochasticity of the simulated biological phenomena for high fidelity purposes. First, we developed a new cell tracking algorithm inspired by motion estimation for video compression that successfully associates cell instances in consecutive frames and can accurately construct dense and deep forests of lineage trees with many cell colonies and generations. Our method can map on the extracted trees cell instance attributes (e.g., cell surface, length, width) and cell life attributes (e.g., cell division time) with high fidelity. The very high bacterial matching accuracy it achieves (98.7%) for complex cell movies exceeds that of prior methods. Our complete, end-to-end cell movies analysis methodology, codenamed BaSCA (Bacterial Single-Cell Analytics), covers cell segmentation, cell tracking, and lineage trees construction for complex cell movies. BaSCA was thoroughly evaluated using datasets generated by different labs and achieved an F-measure rate of 98%. The F-measure remains very high (over 96.7%) even for overcrowded cell movies with many merging colonies and thousands of bacteria in the field of view of the microscope. The information extracted from the cell movie analysis is organized in a relational bio-database, allowing data mining at the single-cell level (single-cell analytics). Another innovative feature is creating different "views" in the data for the direct and friendly visualization of information (visual analytics). In this way, the system allows the user to select subpopulations (colonies, generations, relative cells in the trees) and perform statistical analyses and best distributions parameter estimation. Additionally, this work also contributed to an R package creation codenamed ViSCAR (Visualization and Single-Cell Analytics in R). Using Viscar, one can also correct inevitable segmentation and tracking errors introduced by the image analysis of dense cell movies. Next, we had to enhance the capabilities of CellModeller, an popular open source systems biology tool, to be able to recreate in silico with high fidelity the physical interaction of cells as they grow and divide to form a dense bacterial community, while taking into account the genetic "program" of every individual cell, the underlying stochasticity of cell properties (e.g., division time), the microenvironment conditions, potential cell motion etc. Thus, it became possible to generate in silico experiments of microbial communities growing in two dimensions based on stochastic atomic evolution models for each cell entering the simulation. We created a "digital twin" microbial community prototype to implement the proposed unified strategy for Salmonella's case (Salmonella enterica serovar Typhimurium). To that end, the regulatory networks of gene expression related to the mechanism of intercellular chemical communication (quorum sensing), the mechanism of virulence development of S. typhimurium, and the interaction between the two were integrated into the individual-based cell models we developed. The proof of concept community level "digital twin" simulation model we developed also considers the inherent stochasticity that governs single-cells' growth and division, which affects the patterns of switching their phenotypic state from non-virulent to virulent. Using the digital twin, it is possible to investigate different scenarios in silico. Overall the computational pipeline we have developed can automatically image-analyze cell movies from live-cell microscopy experiments, display efficiently and extract statistical models of single-cell attributes, and utilize them to create realistic synthetic movies. It can also stochastically simulate a microbial community's behavior in space and time. This was demonstrated with a digital twin of a community of pathogen S. typhimurium cells, growing in a microenvironment that promotes stochastic phenotypic switching of the single-cells based on their own personalized genetic network "logic."
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Balomenos, Athanasios
- Μπαλωμένος, Αθανάσιος
Subjects
dc:subject × 30- Time-lapse μικροσκοπία
- Βακτηριακές κυτταρικές ταινίες
- Δέντρα γενεαλογίας
- Δέντρα γενιάς
- Κυτταρομετρία
- Ανάλυση σε επίπεδο μεμονωμένου κυττάρου
- Οπτικοποίηση
- Ρυθμιστικά δίκτυα γονιδίων
- Μοντελοποίηση στοχαστικότητας
- Δυναμική βακτηριακών κοινοτήτων
- Time-lapse microscopy
- Bacterial cell-movies
- Lineage trees
- Generation trees
- Cell cytometry
- Single-cell analytics
- Visualization
- Gene regulatory networks
- Stochasticity modeling
- Bacterial cell community dynamics
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Βιοπληροφορική
- Λογισμικό (software)
- Τεχνητή νοημοσύνη
- Computer and Information Sciences
- Natural Sciences
- Bioinformatics
- Software
- Artificial Intelligence
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/49137