Artificial intelligence exploration of unstable protocells leads to predictable properties and discovery of collective behavior

Edited by Robert H. Austin, Princeton University, Princeton, NJ, and approved December 4, 2017 (received for review June 19, 2017)
January 16, 2018
115 (5) 885-890

Significance

Exploring and understanding the emergence of complex behaviors is difficult even in “simple” chemical systems since the dynamics can rest on a knife edge between stability and instability. Herein, we study the complex dynamics of a simple protocell system, comprising four-component oil droplets in an aqueous environment using an automated platform equipped with artificial intelligence. The system autonomously selects and performs oil-in-water droplet experiments, and then records and classifies the behavior of the droplets using image recognition. The data acquired are then used to build predictive models of the system. Physical properties such as viscosity, surface tension, and density are related to behaviors, as well as to droplet behavioral niches, such as collective swarming.

Abstract

Protocell models are used to investigate how cells might have first assembled on Earth. Some, like oil-in-water droplets, can be seemingly simple models, while able to exhibit complex and unpredictable behaviors. How such simple oil-in-water systems can come together to yield complex and life-like behaviors remains a key question. Herein, we illustrate how the combination of automated experimentation and image processing, physicochemical analysis, and machine learning allows significant advances to be made in understanding the driving forces behind oil-in-water droplet behaviors. Utilizing >7,000 experiments collected using an autonomous robotic platform, we illustrate how smart automation cannot only help with exploration, optimization, and discovery of new behaviors, but can also be core to developing fundamental understanding of such systems. Using this process, we were able to relate droplet formulation to behavior via predicted physical properties, and to identify and predict more occurrences of a rare collective droplet behavior, droplet swarming. Proton NMR spectroscopic and qualitative pH methods enabled us to better understand oil dissolution, chemical change, phase transitions, and droplet and aqueous phase flows, illustrating the utility of the combination of smart-automation and traditional analytical chemistry techniques. We further extended our study for the simultaneous exploration of both the oil and aqueous phases using a robotic platform. Overall, this work shows that the combination of chemistry, robotics, and artificial intelligence enables discovery, prediction, and mechanistic understanding in ways that no one approach could achieve alone.

Continue Reading

Acknowledgments

We thank Dr. Salah Sharabi for assistance with hardware and electronics, Dr. Sergey Zalesskiy for assistance with 1H NMR spectroscopy experiments, Kliment Yanev for providing the base platform, and Dr. Juan Manuel Parrilla Gutierrez. We gratefully acknowledge financial support from the Engineering and Physical Sciences Research Council (EPSRC) for funding (Grants EP/H024107/1, EP/I033459/1, EP/J00135X/1, EP/J015156/1, EP/K021966/1, EP/K023004/1, EP/K038885/1, EP/L015668/1, and EP/L023652/1), the Biotechnology and Biological Sciences Research Council (BBSRC) (Grant BB/M011267/1), and the European Commission (EC) (Projects 610730 EVOPROG and 611640 EVOBLISS). L.C. thanks the Royal Society/Wolfson Foundation for a Merit Award and the European Research Council (ERC) for an Advanced Grant (ERC-ADG, 670467 SMART-POM).

Supporting Information

Supporting Information (PDF)
Appendix (PDF)
Movie S1.
The three-dimensional versions of Fig. 2. The density, surface tension, and dynamic viscosity, predicted as described in SI Appendix, Fig. S5, are plotted for the recipes previously tested, with the color of the datapoint representing the fitness of the droplets in that experiment. This is plotted for the measured (Left) and predicted (Right) droplet fitness, with yellow representing low fitness and red high fitness. From these plots the physical property fitness trends can easily be seen, as can the ability of the model to predict the overall behavioral trends.
Movie S2.
Single-oil and binary-oil formulations dyed with phenolphthalein. In this movie you can see how the different oils behave differently, for example, pentanol goes very pink, has rapid flows, and dissolves, while DEP only goes pink at the interface.
Movie S3.
The same recipes (given in SI Appendix, Table S10) dyed with either Sudan III dye (red, Left, as used for automated experiments) or phenolphthalein (pink at high pH, Right). (i) The droplets can be seen to drift around and seem to interact via white and pink material tethers. This leads to low-activity droplets staying in close proximity and a number of fusion events. It is interesting to note how the pink material is only expelled in distinct directions, with no broader “clouds” being released. (ii) The droplets can be seen to be moving quite smoothly throughout, often in a curved manner punctuated by brief pauses in between movement, often at the edge of the dish. It is interesting to note that there appears to be very little phase mixing––the oil droplets stay clear throughout the indicator movie. The droplets are also seen to bump into one another on several occasions, implying that, in this case, fusion itself is disfavored rather than the close proximity of droplets itself. (iii) The droplets are seen to massively divide in the initial stages of the movie to give many small, unstable droplets. These then drift to the side where they undergo further fusion and division, leading to no active droplets at the end of the movie. It is interesting to see how, in the indicator movie, the droplets appear to influence one another’s movement. The high level of phase mixing is shown by the pink staining of both the oil and aqueous phases in the indicator movie and by the pink coloration in the dye movie due to deprotonation of the Sudan III dye. (iv) Low movement is observed in this movie, as for the most part all of the droplets are stuck to the edge of the dish. We also see the gentle expulsion of material from the indicator droplets and slow flows in the droplets which influence their movement. (v) For this low division fitness recipe, we observe the droplets moving straight to the edge of the dish, thus giving an active droplet count of 0. Interestingly, when at the edge of the dish, the droplets appear to wobble and expel material––showing how the apparently inactive droplets of the dye movie are actually actively expelling material. (vi) In this movie we see classic swarming behavior. Initially, we have massive division, which is followed by the collective movement of the small droplets and their rapid fusion to form larger droplets, before they stick to the side. It is interesting to see how the droplets appear to interact via their pink clouds, often following the same path to the edge of the dish. It is important to note that fusion does not always occur as rapidly for swarming as it does in this movie. (vii) In this movie, the droplets are initially relatively inactive before they begin to move in fairly rapid spurts. It is apparent from the indicator movie that, during this initial lower-activity phase, the droplets are actually expelling material. It could be that once this material is expelled the oil phase is more optimal for movement, or that the dissolution of the oil in the aqueous influences the surface tension such that movement is promoted. For example, the material could be inhomogeneously dissolved, leading to the surface tension variations and the promotion of Marangoni instabilities, leading to the variable droplet movement observed. (viii) In this movie the droplets are seen to be stationary, stuck to the side. There appears to be no oil–aqueous phase mixing, as evidenced by the lack of any pink coloration.
Movie S4.
Operation of the platform.
Movie S5.
Observed droplet behaviors: movement, division, vibration, fusion, pulsing, and swarming. Movies correspond to the recipes shown in SI Appendix, Table S11.
Movie S6.
Highest movement formulations from each genetic algorithm run: oil only, aqueous only, and simultaneous aqueous- and oil-phase optimization. Movies correspond to the recipes shown in SI Appendix, Table S12.

References

1
S Lach, SM Yoon, BA Grzybowski, Tactic, reactive, and functional droplets outside of equilibrium. Chem Soc Rev 45, 4766–4796 (2016).
2
JMP Gutierrez, T Hinkley, JW Taylor, K Yanev, L Cronin, Evolution of oil droplets in a chemorobotic platform. Nat Commun 5, 5571 (2014).
3
MM Hanczyc, T Toyota, T Ikegami, N Packard, T Sugawara, Fatty acid chemistry at the oil-water interface: Self-propelled oil droplets. J Am Chem Soc 129, 9386–9391 (2007).
4
I Lagzi, S Soh, PJ Wesson, KP Browne, BA Grzybowski, Maze solving by chemotactic droplets. J Am Chem Soc 132, 1198–1199 (2010).
5
AJ Dzieciol, S Mann, Designs for life: Protocell models in the laboratory. Chem Soc Rev 41, 79–85 (2012).
6
M Schmitt, H Stark, Marangoni flow at droplet interfaces: Three-dimensional solution and applications. Phys Fluids 28, 12106 (2016).
7
AJ Bissette, B Odell, SP Fletcher, Physical autocatalysis driven by a bond-forming thiol-ene reaction. Nat Commun 5, 4607 (2014).
8
J Ortega-Arroyo, AJ Bissette, P Kukura, SP Fletcher, Visualization of the spontaneous emergence of a complex, dynamic, and autocatalytic system. Proc Natl Acad Sci USA 113, 11122–11126 (2016).
9
I Bottero, J Huck, T Kosikova, D Philp, A synthetic replicator drives a propagating reaction–diffusion front. J Am Chem Soc 138, 6723–6726 (2016).
10
F Haudin, JHE Cartwright, F Brau, A De Wit, Spiral precipitation patterns in confined chemical gardens. Proc Natl Acad Sci USA 111, 17363–17367 (2014).
11
LM Barge, et al., From chemical gardens to chemobrionics. Chem Rev 115, 8652–8703 (2015).
12
Y Qiao, M Li, R Booth, S Mann, Predatory behaviour in synthetic protocell communities. Nat Chem 9, 110–119 (2017).
13
AE Engelhart, KP Adamala, JW Szostak, A simple physical mechanism enables homeostasis in primitive cells. Nat Chem 8, 448–453 (2016).
14
J Adler, W-W Tso, “Decision”-making in bacteria: Chemotactic response of Escherichia coli to conflicting stimuli. Science 184, 1292–1294 (1974).
15
PJM Van Haastert, PN Devreotes, Chemotaxis: Signalling the way forward. Nat Rev Mol Cell Biol 5, 626–634 (2004).
16
T Toyota, N Maru, MM Hanczyc, T Ikegami, T Sugawara, Self-propelled oil droplets consuming “fuel” surfactant. J Am Chem Soc 131, 5012–5013 (2009).
17
S Miura, et al., pH-induced motion control of self-propelled oil droplets using a hydrolyzable gemini cationic surfactant. Langmuir 30, 7977–7985 (2014).
18
J Cejková, M Novák, F Stěpánek, MM Hanczyc, Dynamics of chemotactic droplets in salt concentration gradients. Langmuir 30, 11937–11944 (2014).
19
LE Scriven, CV Sternling, The Marangoni effects. Nature 187, 186–188 (1960).
20
AE Eiben, S Kernbach, E Haasdijk, Embodied artificial evolution: Artificial evolutionary systems in the 21st century. Evol Intell 5, 261–272 (2012).
21
SL Harding, JF Miller, EA Rietman, Evolution in materio: Exploiting the physics of materials for computation. Int J Unconv Comput 4, 155–194 (2008).
22
GS Hornby, JD Lohn, DS Linden, Computer-automated evolution of an X-band antenna for NASA’s Space Technology 5 mission. Evol Comput 19, 1–23 (2011).
23
AE Eiben, J Smith, From evolutionary computation to the evolution of things. Nature 521, 476–482 (2015).
24
SJ Gould, Darwinism and the expansion of evolutionary theory. Science 216, 380–387 (1982).
25
MI Jordan, TM Mitchell, Machine learning: Trends, perspectives, and prospects. Science 349, 255–260 (2015).
26
P Raccuglia, et al., Machine-learning-assisted materials discovery using failed experiments. Nature 533, 73–76 (2016).
27
G Carleo, M Troyer, Solving the quantum many-body problem with artificial neural networks. Science 355, 602–606 (2017).
28
JS Smith, O Isayev, AE Roitberg, ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost. Chem Sci (Camb) 8, 3192–3203 (2017).
29
CA Schneider, WS Rasband, KW Eliceiri, NIH Image to ImageJ: 25 years of image analysis. Nat Methods 9, 671–675 (2012).
30
J-C Baret, Surfactants in droplet-based microfluidics. Lab Chip 12, 422–433 (2012).
31
F Caschera, S Rasmussen, MM Hanczyc, An oil droplet division–fusion cycle. ChemPlusChem 78, 52–54 (2013).
32
J Lehman, KO Stanley, Exploiting open-endedness to solve problems through the search for novelty. Proceedings of the Eleventh International Conference on Artificial Life (ALIFE XI) (MIT Press, Cambridge, MA), pp. 329–336 (2008).
33
PY Oudeyer, F Kaplan, VV Hafner, Intrinsic motivation systems for autonomous mental development. IEEE Trans Evol Comput 11, 265–286 (2007).

Information & Authors

Information

Published in

Go to Proceedings of the National Academy of Sciences
Go to Proceedings of the National Academy of Sciences
Proceedings of the National Academy of Sciences
Vol. 115 | No. 5
January 30, 2018
PubMed: 29339510

Classifications

Submission history

Published online: January 16, 2018
Published in issue: January 30, 2018

Keywords

  1. artificial intelligence
  2. protocell models
  3. complex chemical systems
  4. emergence
  5. machine learning

Acknowledgments

We thank Dr. Salah Sharabi for assistance with hardware and electronics, Dr. Sergey Zalesskiy for assistance with 1H NMR spectroscopy experiments, Kliment Yanev for providing the base platform, and Dr. Juan Manuel Parrilla Gutierrez. We gratefully acknowledge financial support from the Engineering and Physical Sciences Research Council (EPSRC) for funding (Grants EP/H024107/1, EP/I033459/1, EP/J00135X/1, EP/J015156/1, EP/K021966/1, EP/K023004/1, EP/K038885/1, EP/L015668/1, and EP/L023652/1), the Biotechnology and Biological Sciences Research Council (BBSRC) (Grant BB/M011267/1), and the European Commission (EC) (Projects 610730 EVOPROG and 611640 EVOBLISS). L.C. thanks the Royal Society/Wolfson Foundation for a Merit Award and the European Research Council (ERC) for an Advanced Grant (ERC-ADG, 670467 SMART-POM).

Notes

This article is a PNAS Direct Submission.

Authors

Affiliations

Laurie J. Points
WestCHEM, School of Chemistry, University of Glasgow, Glasgow G12 8QQ, United Kingdom
James Ward Taylor
WestCHEM, School of Chemistry, University of Glasgow, Glasgow G12 8QQ, United Kingdom
Jonathan Grizou
WestCHEM, School of Chemistry, University of Glasgow, Glasgow G12 8QQ, United Kingdom
Kevin Donkers
WestCHEM, School of Chemistry, University of Glasgow, Glasgow G12 8QQ, United Kingdom
WestCHEM, School of Chemistry, University of Glasgow, Glasgow G12 8QQ, United Kingdom

Notes

1
To whom correspondence should be addressed. Email: [email protected].
Author contributions: J.G. and L.C. designed research; L.J.P., J.W.T., J.G., and K.D. performed research; L.J.P., J.W.T., and K.D. contributed new reagents/analytic tools; L.J.P., J.W.T., and J.G. analyzed data; and L.J.P., J.G., and L.C. wrote the paper.

Competing Interests

The authors declare no conflict of interest.

Metrics & Citations

Metrics

Note: The article usage is presented with a three- to four-day delay and will update daily once available. Due to ths delay, usage data will not appear immediately following publication. Citation information is sourced from Crossref Cited-by service.


Citation statements

Altmetrics

Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

Cited by

    Loading...

    View Options

    View options

    PDF format

    Download this article as a PDF file

    DOWNLOAD PDF

    Get Access

    Login options

    Check if you have access through your login credentials or your institution to get full access on this article.

    Personal login Institutional Login

    Recommend to a librarian

    Recommend PNAS to a Librarian

    Purchase options

    Purchase this article to get full access to it.

    Single Article Purchase

    Artificial intelligence exploration of unstable protocells leads to predictable properties and discovery of collective behavior
    Proceedings of the National Academy of Sciences
    • Vol. 115
    • No. 5
    • pp. 823-E1075

    Media

    Figures

    Tables

    Other

    Share

    Share

    Share article link

    Share on social media