Machine learning based analytics for bacteria cell cycle characterization using super resolution microscopy

Project Information

BALTIC

Grant agreement ID: 890169

DOI

10.3030/890169

Project closed

EC signature date 20 March 2020

Start date 1 April 2020

End date 7 October 2022

Funded under

• EXCELLENT SCIENCE - Marie Skłodowska-Curie Actions

Total cost

€ 191 149,44

EU contribution

€ 191 149,44

Coordinated by

ECOLE POLYTECHNIQUE FEDERALE DE LAUSANNE
Switzerland

Periodic Reporting for period 1 - BALTIC (Machine learning based analytics for bacteria cell cycle characterization using super resolution microscopy)

Reporting period: 2020-04-01 to 2022-03-31

Summary of the context and overall objectives of the project

Super resolution microscopy provides an unprecedented insight into the molecular organization in cells. Whilst retaining molecular specificity and multiplexing, it generates images with near molecular resolution in the case of single molecule localisation microscopy (SMLM), typically 10-30nm. This gain in resolution however comes at a cost: complex sample preparation and, for the most part, loss of dynamic information (i.e. fixed cells). This research project aimed at tackling some of these limitations for the study of bacteria cell cycle (BCC), a fundamental dynamic process with clinical applications (e.g. bacterial infection, drug targeting). SMLM enables to visualize and investigate a mixed population of fixed cells (both cell wall and specific protein markers) at various cell stages at the nanoscale. We aimed at retrieving from these static images the otherwise lost BCC. Existing approaches for such "pseudo time" analysis, also called ordering, often rely heavily on pre-designed models and/or multiple protein markers. The significant drawback of such approaches is that they are heavily user-biased, cell specific and using most of the available resources the microscope has to offer. As a result, they remain constrained for the most part to cell specific studies, focusing primarily, or only, on characterizing the cell cycle, rather than providing a general framework for more complex and scalable studies. What I proposed instead is to rely on generative models, a subclass of unsupervised machine leaning tools, to infer hidden information, including the BCC, directly from the images. Generative models, including variational auto encoders (VAE) such as implemented for this work, are designed to learn how to best replicate the inputted data, in this case images of segmented bacteria. VAE have the particularity to do so by compressing the inputted images to a very low dimensional space or latent space. In this latent space, only the essential information is retained, which includes the BCC along which the collected images/cells get ordered in an automated fashion. It requires no a priori knowledge and, as such, uniquely allows for simultaneous multi cell types studies. Such data driven approach had a significant impact on our understanding of complex data sets such as genomics, as it provides a unique high information content representation of the inputted data. We foresee that it will play a similar role: enabling super resolution microscopy to become a routinely used tool to address fundamental and complex biological issues.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

Imaging: various bacteria types were imaged using a home built SMLM as well as commercial SR set ups, unravelling their cell shape at the nanoscale as well as the relocation of key protein markers’ organisation along the BCC.
Analysis tools:
1: both segmentation and structural characterisation tools have been developed. For segmentation two approaches were followed. Firstly, an SMLM specific approach relying on relative assessment of the local density of points for the segmentation of individual bacteria from the initial field of view (FOV), followed by tessellation to extract the underlying skeleton of the cell shape. The second approach attempts at providing a generalisable solution to segmentation with intensity thresholding and connectivity. Both techniques were thoroughfully validated with simulated and experimental data sets.
2: I relied on generative models, a subgroup of unsupervised machine learning, for the extraction of dynamic information from fixed cells (i.e. static) images. It provides the considerable advantage to require no a priori knowledge or pre-defined model, and, hence, is truly compatible with multi cell types studies. I developed for the most part, data specific auto encoders. They take the segmented images as input and learn how to replicate these images. The architecture allows in the process to reduce inputted images to a very low dimensional space from which dynamic information, and more, can be easily extracted.

Dissemination:
1;List the conferences attended (all of which included reference to EU funding):
- SMLMS (2020) - online - organization of a track
- SMLMS (2021) – Switzerland - in person - organization of a discussion & challenge around clustering for SMLM
- LS2 webinar series (over 2021) - online - organization & board member
- MiFoBio (2022) - France - in person - track organization/chair & invited speaker
- COMPARE seminar series (2022) - UK - in person - invited speaker
- Theory of Living Matter Group seminar (2022) - UK - in person - invited speaker
- SciLifeLab annual meeting (2022)- Sweden - in person - invited speaker

2: List of publications over the duration of my MSCA fellowship:
In preparation
- J. Griffié et al. An unsupervised approach to extract dynamic information from static super-resolved images. To be submitted in December 2022 to Nature Machine intelligence.
Submitted work
- J. Griffié, T. Pham, C. Sieben, R. Lang, V. Cevher, S. Holden, M. Unser, S. Manley, D. Sage. Virtual-SMLM, a virtual environment for real-time interactive SMLM acquisition. BioRxiv (2020), revisions for submission at Nature Communications.
Accepted & published work
- M. Lelek, M.T. Gyparaki, G. Beliu, F. Schueder, J. Griffié, S. Manley, R. Jungmann, M. Sauer, M. Lakadamyali, C. Zimmer. Single-molecule localization microscopy. Nature Reviews Methods Primers(2022).
- D. Mahecic, W.L. Stepp, C. Zhang, J. Griffié, M. Weigert, S. Manley. Event-driven acquisition for content-enriched microscopy. Nature Methods (2022).
- D.J. Nieves, J.A. Pike, F. Levet, J. Griffié, D. Sage, E.A.K. Cohen, J.B. Sibarita, M. Heilemann, D.M. Owen. A framework for evaluating the performance of SMLM cluster analysis algorithms. Nature Methods (2022).
- C. Zhang, L. Reymond, O. Rutschmann, M.A. Meyer, J. Denereaz, J. Qiao, F. Ryckebusch, J. Griffié, W.L. Stepp, S. Manley. Fluorescent d-Amino Acids for Super-resolution Microscopy of the Bacterial Cell Wall. ACS Chemical Biology (2022).
- L.G. Jensen, T.Y. Hoh, D.J. Williamson, J. Griffié, D. Sage, P. Rubin-Delanchy, D.M. Owen. Correction of multiple-blinking artefacts in photoactivated localisation microscopy. Accepted at Nature Methods (2022).

All accepted & submitted work have been made accessible on bioArchive (open access) prior to their publication or were published in open access journals.

Exploitation:
I provided a detailed data management plan which explained where: the raw data, the models and the meta data are made available. All are open access.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

The project has led to a novel methodology for the study of BCC. Combining super resolved images and VAE, we were able to infer lost dynamic information at the nanoscale, as well as provide a unique framework with no a priori knowledge requirements in order to address more complex biological issues, including multi cell types studies.

Unsupervised cell ordering