Computational tools for designing modular biosystems

Computational tools for designing modular biosystems

My slides for Juliacon 2020.


Michiel Stock

July 29, 2020



    on Unsplash Michiel Stock @michielstock michielfmstock@gmailcom KERMIT 1
  2. Modular design in biotech ( , , ) set sequence

    graph composition Combine modules in a design Build & test building blocks “modules” … e.g., drug formulations e.g., plant ecosystems e.g., modular protein e.g., microbial communities health biotech ecology 2
  3. 0 2 4 6 0.0 x 0 2 4 6

    -2 -1 0 1 2 3 0 2 4 6 0.000 0.025 0.050 0.075 0.100 x f(x) R = 4 Bayesian optimization for modular design Gaussian process with prediction intervals acquisition function to determining the next designs to test next design to build and test true objective (unknown) 3
  4. Designing co-cultures to degrade BAM t P t P co-culture

    incubation α x x x x mineralization curve score new cultures with acquisition function mineralization curve prediction
 with uncertainty estimates Gompertz fitting GP fitting select most promising 
 co-culture for screening new
 co-culture 2,6-dichlorobenzamide (toxic) finding the co-culture much faster using Bayesian optimization Daly, A. J., Stock, M., Baetens, J. M., & De Baets, B. (2019). Guiding mineralization co-culture discovery using Bayesian optimization. Environmental Science and Technology, 53(24), 14459–14469. 4
  5. Finding new enzybiotics PhD project Bjorn Criel, joint work with

    Yves Briers’ group Finding more efficient proteins to kill bacteria Package in development: 5
  6. High-throughput screenings SCREENING Combinatorially generated population Subpopulation with desired properties

    Find common characteristics 6
  7. Plant phenotyping Golden Gate Cloning to randomly knock out genes

    in plants package in development 7
  8. ‘Natural’ experiments Naturally occurring designs ( OR ) AND design

    rules Rules or models for generating good designs 8
  9. Finding design rules in natural phage lytic proteins 9

  10. Cocktail composition simple linear model to determine ratios of mixer

    based on co-occurrence mixers added log mixer fractions 10
  11. Why Julia? •Speed (duh) •Great ecosystem of scientific libraries… •…that

    play well together (e.g., autodiff everything!) •Relatively easy to write interfaces and publish packages 11
  12. People involved Phages & enzybiotics: • Yves Briers • Dimitri

    Boeckaerts • Bernard De Baets • Sam Dierickx • Hans Gerstmans • Bjorn Criel • Steff Taelman Mixology: • Bernard De Baets • Julie Vermout Co-culture prediction: • Aisling Daly • Bernard De Baets Plant knockout phenotyping: • Rafael Buono • Bernard De Baets • Thomas Jacobs • Tom Ruttink 12