Learning dynamic image representations for self-supervised cell cycle annotation

Ulicna, Kristina; Kelkar, Manasi; Soelistyo, Christopher J; Charras, Guillaume; Lowe, Alan R.

doi:10.1101/2023.05.30.542796

Cited by 3 publications

(2 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While it is theoretically feasible to discern each embryonic stage ranging from a single nucleus to 558 nuclei at the conclusion of embryogenesis, such an undertaking would be exceedingly challenging for a high-throughput approach [38][39][40] or require staining of additional markers [41]. Therefore, we employ an autoencoder-based image classifier to learn representations from fixed, DAPI-stained C. elegans embryos to automatically stage them in an approach similar to [42]. We rely on a manually created training dataset for each developmental stage to achieve reliable predictions from the learned representations.…”

Section: Computational Staging Of C Elegans Embryos and Smfish Analysismentioning

confidence: 99%

Analysis of developmental gene expression using smFISH andin silicostaging ofC. elegansembryos

Breimann,

Bahry,

Zouinkhi

et al. 2024

Preprint

View full text Add to dashboard Cite

Regulation of transcription during embryogenesis is key to development and differentiation. To study transcript expression throughout Caenorhabditis elegans embryogenesis at single-molecule resolution, we developed a high-throughput single-molecule fluorescence in situ hybridization (smFISH) method that relies on computational methods to developmentally stage embryos and quantify individual mRNA molecules in single embryos. We applied our system to sdc-2, a zygotically transcribed gene essential for hermaphrodite development and dosage compensation. We found that sdc-2 is rapidly activated during early embryogenesis by increasing both the number of mRNAs produced per transcription site and the frequency of sites engaged in transcription. Knockdown of sdc-2 and dpy-27, a subunit of the dosage compensation complex (DCC), increased the number of active transcription sites for the X chromosomal gene dpy-23 but not the autosomal gene mdh-1, suggesting that the DCC reduces the frequency of dpy-23 transcription. The temporal resolution from in silico staging of embryos showed that the deletion of a single DCC recruitment element near the dpy-23 gene causes higher dpy-23 mRNA expression after the start of dosage compensation, which could not be resolved using mRNAseq from mixed-stage embryos. In summary, we have established a computational approach to quantify temporal regulation of transcription throughout C. elegans embryogenesis and demonstrated its potential to provide new insights into developmental gene regulation.

show abstract

Section: Computational Staging Of C Elegans Embryos and Smfish Analysismentioning

confidence: 99%

Analysis of developmental gene expression using smFISH andin silicostaging ofC. elegansembryos

Breimann,

Bahry,

Zouinkhi

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…[17][18][19] Scientists have also applied deep learning to microscopy images directly to predict single-cell phenotypes (reviewed in Pratapa et al 20 ), most often using convolutional neural networks 21 or autoencoders. 22 However, these approaches do not rigorously test the generalizability of single-cell phenotype prediction in new datasets. Other approaches have successfully mapped bulk signatures across datasets, but these primarily focus on linking perturbation signatures rather than individual single-cell phenotypes.…”

Section: Introductionmentioning

confidence: 99%

Toward generalizable phenotype prediction from single-cell morphology representations

Tomkinson,

Kern,

Mattson

et al. 2024

Preprint

View full text Add to dashboard Cite

Functional cell processes (e.g., molecular signaling, response to environmental stimuli, mitosis, etc.) impact cell phenotypes, which scientists can easily and robustly measure with cell morphology. However, linking these morphology measurements with phenotypes remains challenging because biologically interpretable phenotypes require manually annotated labels. Automatic phenotype annotation from cell morphology would link biological processes with their phenotypic outcomes and deepen understanding of cell function. We propose that nuclear morphology can be a predictive marker for cell phenotypes that is generalizable across cell types. Nucleus morphology is commonly and easily accessible with microscopy, but annotating specific phenotypic information requires labels. Therefore, we reanalyzed a pre-labeled, publicly-available nucleus microscopy dataset from the MitoCheck consortium to predict single-cell phenotypes. We extracted single-cell morphology features using CellProfiler and DeepProfiler, which provide fast, robust, and generalizable data processing pipelines. We trained multinomial, multi-class elastic net logistic regression models to classify nuclei into one of 15 phenotypes such as "Anaphase", "Apoptosis", and "Binuclear". In a held-out test set, we observed an overall F1 score of 0.84, where individual phenotype scores ranged from 0.64 (indicating moderate performance) to 0.99 (indicating high performance). Notably, phenotypes such as "Elongated", "Metaphase", and "Apoptosis" showed high performance. While CellProfiler and DeepProfiler morphology features were generally equally effective, combining feature spaces yielded the best results for 9 of the 15 phenotypes. However, leave-one-image-out (LOIO) cross-validation analysis showed a significant performance decline, indicating our model could not reliably predict phenotype in new single images. Poor performance, which we show was unrelated to factors like illumination correction or model selection, limits generalizability to new datasets and highlights the challenges of morphology to phenotype annotation. Nevertheless, we modified and applied our approach to the JUMP Cell Painting pilot data. Our modified approach improved dataset alignment and highlighted many perturbations that are known to be associated with specific phenotypes. We propose several strategies that could pave the way for more generalizable methods in single-cell phenotype prediction, which is a step toward morphology representation ontologies that would aid in cross-dataset interpretability.

show abstract

Orientation-invariant autoencoders learn robust representations for shape profiling of cells and organelles

Burgess,

Nirschl,

Zanellati

et al. 2024

Nat Commun

View full text Add to dashboard Cite

Cell and organelle shape are driven by diverse genetic and environmental factors and thus accurate quantification of cellular morphology is essential to experimental cell biology. Autoencoders are a popular tool for unsupervised biological image analysis because they learn a low-dimensional representation that maps images to feature vectors to generate a semantically meaningful embedding space of morphological variation. The learned feature vectors can also be used for clustering, dimensionality reduction, outlier detection, and supervised learning problems. Shape properties do not change with orientation, and thus we argue that representation learning methods should encode this orientation invariance. We show that conventional autoencoders are sensitive to orientation, which can lead to suboptimal performance on downstream tasks. To address this, we develop O2-variational autoencoder (O2-VAE), an unsupervised method that learns robust, orientation-invariant representations. We use O2-VAE to discover morphology subgroups in segmented cells and mitochondria, detect outlier cells, and rapidly characterise cellular shape and texture in large datasets, including in a newly generated synthetic benchmark.

show abstract

Learning dynamic image representations for self-supervised cell cycle annotation

Cited by 3 publications

References 17 publications

Analysis of developmental gene expression using smFISH andin silicostaging ofC. elegansembryos

Analysis of developmental gene expression using smFISH andin silicostaging ofC. elegansembryos

Toward generalizable phenotype prediction from single-cell morphology representations

Orientation-invariant autoencoders learn robust representations for shape profiling of cells and organelles

Contact Info

Product

Resources

About