Molecular recording of mammalian embryogenesis

Chan, M.K.; Smith, Zachary D.; Grosswendt, Stefanie; Kretzmer, Helene; Norman, Thomas M.; Adamson, Britt; Jost, Marco; Quinn, Jeffrey J.; Yang, Dian; Jones, Matthew G.; Khodaverdian, Alex; Yosef, Nir; Meissner, Alexander; Weissman, Jonathan S.

doi:10.1038/s41586-019-1184-5

Cited by 278 publications

(425 citation statements)

References 45 publications

(34 reference statements)

Supporting

Mentioning

411

Contrasting

Order By: Relevance

“…Such labels might be functional responses to drug treatment, gene perturbation or other stimulations, which may not trivially be visible on expression clusters. For cell lines, retrieving such information requires creative experimental methods such as traceable genetic manipulation 22,23 , which allows tracing cell fates over several generations. In tissue samples, the transcriptome of cells depends on its direct neighbourhood 24,25 , i.e.…”

Section: Discussionmentioning

confidence: 99%

Deep learning does not outperform classical machine learning for cell-type annotation

Büttner

Andriamanga

et al. 2019

Preprint

View full text Add to dashboard Cite

Deep learning has revolutionized image analysis and natural language processing with remarkable accuracies in prediction tasks, such as image labeling or word identification. The origin of this revolution was arguably the deep learning approach by the Hinton lab in 2012, which halved the error rate of existing classifiers in the then 2year-old ImageNet database 1 . In hindsight, the combination of algorithmic and hardware advances with the appearance of large and well-labeled datasets has led up to this seminal contribution. The emergence of large amounts of data from single-cell RNA-seq and the recent global effort to chart all cell types in the Human Cell Atlas has attracted an interest in deep-learning applications. However, all current approaches are unsupervised, i.e., learning of latent spaces without using any cell labels, even though supervised learning approaches are often more powerful in feature learning and the most popular approach in the current AI revolution by far. Here, we ask why this is the case. In particular we ask whether supervised deep learning can be used for cell annotation, i.e. to predict cell-type labels from single-cell gene expression profiles. After evaluating 6 classification methods across 14 datasets, we notably find that deep learning does not outperform classical machine-learning methods in the task. Thus, cell-type prediction based on gene-signature derived celltype labels is potentially too simplistic a task for complex non-linear methods, which demands better labels of functional single-cell readouts. We, therefore, are still waiting for the "ImageNet moment" in single-cell genomics.

show abstract

Section: Discussionmentioning

confidence: 99%

Deep learning does not outperform classical machine learning for cell-type annotation

Büttner

Andriamanga

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…To investigate how 132 close the cellular states identified in TLS resemble those in embryos, we mapped our single-cell 133 transcriptomes to a scRNA-seq compendium of post-occipital embryonic cellular subtypes (Fig. 134 4a) 35 . The data revealed globally a high accordance of TLS and embryonic cell types including 135 characteristic marker genes and pairwise comparison of mapped clusters identified only a small 136 fraction of differentially expressed genes (Fig.…”

mentioning

confidence: 99%

Mouse embryonic stem cells self-organize into trunk-like structures with neural tube and somites

Veenvliet

Bolondi

Kretzmer

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

One sentence summary: A platform for generating trunk-like-structures with precursors of spinal cord, bone and muscle from stem cells in a dish Abstract: Post-implantation embryogenesis is a highly dynamic process comprising multiple lineage decisions and morphogenetic changes inaccessible to deep analysis in vivo. Mouse embryonic stem cells (mESCs) can form aggregates reflecting the post-occipital embryo (gastruloids), but lacking proper morphogenesis. Here we show that embedding of aggregates derived from mESCs in an extracellular matrix compound results in Trunk-Like-Structures (TLS) with a high level of organization comprising a neural tube and somites. Comparative single-cell RNA-seq analysis demonstrates that TLS execute gene-regulatory programs in an embryo-like order, and generate primordial germ cell like cells (PGCLCs). TLS lacking Tbx6 form ectopic neural tubes, mirroring the embryonic mutant phenotype. ESC-derived trunk-like structures thus constitute a novel powerful in vitro platform for investigating lineage decisions and morphogenetic processes shaping the post-implantation embryo.2 Vertebrate post-implantation development comprises a multitude of complex morphogenetic 1 processes, which result from self-organization of stem cells and their descendants shaping the 2 embryonic body plan 1 . Recently developed stem cell models represent powerful platforms for 3 deconstructing the dynamics of these processes that are inaccessible in the embryo 1, 2 . The 4 most advanced models in terms of developmental stage accomplished so far are gastruloids 2,3 , 5 aggregates of mESCs able to self-organize into the three body axes 3 . However, gastruloids lack 6 proper morphogenesis, such as a failure of neural tube formation from neural cells or somite 7 condensation from presomitic cells 3 . In vivo, the extracellular matrix (ECM) has a critical role in 8 tissue morphogenesis 4 . In vitro, matrigel can serve as ECM surrogate, and culture media 9 supplemented with a low percentage of matrigel have been shown to induce complex 10 morphogenesis in organoids 5 . 11We therefore explored if embryo-like morphogenetic features could be induced by 12 embedding 96h post aggregation gastruloids in 5% matrigel for an additional 24h period ( Fig. 13 1a). We also tested a combination of matrigel together with the WNT signaling activator 14 CHIR99021 (CHIR) and the BMP signaling inhibitor LDN193189 (LDN), that have been reported 15 to induce a (pre-)somitic mesoderm ((P)SM) fate in 2D and 3D differentiation protocols 6-8 . To 16 facilitate high-throughput characterization and quantification of our conditions, we generated 17 mESCs with T::H2B-mCherry (hereafter T mCH ) and Sox2::H2B-Venus (hereafter Sox2 VE ) reporters, 18 marking the mesodermal (ME) or neural (NE) lineage respectively (Extended Data Fig. 1a). 19 Strikingly, embedding in 5% matrigel alone was sufficient for segmentation in the T mCH+ ME 20 domain and formation of a Sox2 VE+ neural tube like structure ( Fig. 1b,d; Extended Data Fig. 21 1b,c). The vast majorit...

show abstract

“…One of the critical questions in life science is how embryo becomes a complex multicellular organism. 24,25 The top-down approach generally refers to the use of natural embryos as research objects to evaluate embryonic development ( Figure 1). 17 In the early stage, mouse oocytes are fertilized and divided into the oviducts.…”

Section: Top-downmentioning

confidence: 99%

Bioengineered microenvironment to culture early embryos

Guo

Wang

et al. 2020

Cell Proliferation

View full text Add to dashboard Cite

The abnormalities of early post‐implantation embryos can lead to early pregnancy loss and many other syndromes. However, it is hard to study embryos after implantation due to the limited accessibility. The success of embryo culture in vitro can avoid the challenges of embryonic development in vivo and provide a powerful research platform for research in developmental biology. The biophysical and chemical cues of the microenvironments impart significant spatiotemporal effects on embryonic development. Here, we summarize the main strategies which enable researchers to grow embryos outside of the body while overcoming the implantation barrier, highlight the roles of engineered microenvironments in regulating early embryonic development, and finally discuss the future challenges and new insights of early embryo culture.

show abstract

Molecular recording of mammalian embryogenesis

Cited by 278 publications

References 45 publications

Deep learning does not outperform classical machine learning for cell-type annotation

Deep learning does not outperform classical machine learning for cell-type annotation

Mouse embryonic stem cells self-organize into trunk-like structures with neural tube and somites

Bioengineered microenvironment to culture early embryos

Contact Info

Product

Resources

About