Modelling the Failure of Microalgae Production

The overall goal of WP3 is to analyse the mechanisms underlying the productivity decreases in microalgal PBRs, identify early warning signals and evaluate, from the whole set of identified signals, those that meet the specified criteria of scalability, reliability, and affordability.

• Task 3.1. Linking community structure, gene expression and bioreactor functioning through state transitions
  This task will integrate all the accumulated information on microbial community structure (metabarcoding), functional potential (metagenomics), gene expression (metatranscriptomics), and expressed functions (metaproteomics) from Task 3 (work package 1), as well as chemical fingerprinting data from Task 4 (work package 1). MetaOMICs data tables from perturbation experiments will be linked with data tables containing routine measurements and chemical fingerprint analyses. Metabolic reconstruction and functional pathways will be generated and extracted for all MAGs, gene expression patterns, and expressed proteins belonging to these will be quantified across the PBR transitions. Key features (taxa, mechanisms, function, interactions) correlating with bioreactor transitions will be identified through a combination weighted correlation network analysis, multivariate analysis, eigengene networks, and Sankey diagrams.
• Task 3.2. Modelling photobioreactor systems
  This task aims to i) develop a new plankton ecosystem model, based on the MIT DARWIN ecosystem model and the approach of Thingstad (see reference in methodology) to simulate mortality caused by pathogenic microorganisms, and ii) to run model simulations in order to explore the dynamics of microalgae under multiple perturbation scenarios. The model will be solved in a simple chemostat photobioreactor setup. The MIT DARWIN model of this chemostat bioreactor setup, written in an open-source programming language (GNU OCTAVE), will be freely available for download in a GitHub repository (under creative commons copy-left). All simulations will be intended to reproduce the results of the experimental Tasks 1 and 2 (work package 1).
• Task 3.3 Empirical dynamic modelling for early warning signals identification
  Time series data of chemical and biological data will be assembled and analysed using the Empirical Dynamic Modelling (EDM) framework described in section 1.3b Methodology (subsection AS2.3 Early warning signals identification) of Annex 1 Part B. EDM will compute from time-series data interaction network on each time step; the topology and the strength of interactions. The time-varying interaction network provides an estimate of the stability of the system over time and allows identifying the main contributor/s to the process failure. Massive data acquisition using metabarcoding (Task 1.3) and chemical fingerprinting (Task 1.4), along with routine measurements of system parameters (Tasks 1.1 and 1.2), will provide i) an unprecedented analysis of interaction networks (the interactomes) in bioreactors, and ii) efficient early warning signals for process failure in algal photobioreactors. A qualitative analysis of warning indicators will be carried out to evaluate, from all identified warnings, the best candidates according to the criteria of scalability, ease of determination (measurement), and affordability (cost of measurement). The reliability of the early warning indicators will be tested by repeating the analyses on time series derived from experimental replicates. Warning thresholds and warning times will be also calculated from the EDM results.