MELLODDY: Cross-pharma Federated Learning at Unprecedented Scale Unlocks Benefits in QSAR without Compromising Proprietary Information.

Federated multipartner machine learning has been touted as an appealing and efficient method to increase the effective training data volume and thereby the predictivity of models, particularly when the generation of training data is resource-intensive. In the landmark MELLODDY project, indeed, each of ten pharmaceutical companies realized aggregated improvements on its own classification or regression models through federated learning. To this end, they leveraged a novel implementation extending multitask learning across partners, on a platform audited for privacy and security. The experiments involved an unprecedented cross-pharma data set of 2.6+ billion confidential experimental activity data points, documenting 21+ million physical small molecules and 40+ thousand assays in on-target and secondary pharmacodynamics and pharmacokinetics. Appropriate complementary metrics were developed to evaluate the predictive performance in the federated setting. In addition to predictive performance increases in labeled space, the results point toward an extended applicability domain in federated learning. Increases in collective training data volume, including by means of auxiliary data resulting from single concentration high-throughput and imaging assays, continued to boost predictive performance, albeit with a saturating return. Markedly higher improvements were observed for the pharmacokinetics and safety panel assay-based task subsets.

Asymmetric division of contractile domains couples cell positioning and fate specification.

During pre-implantation development, the mammalian embryo self-organizes into the blastocyst, which consists of an epithelial layer encapsulating the inner-cell mass (ICM) giving rise to all embryonic tissues. In mice, oriented cell division, apicobasal polarity and actomyosin contractility are thought to contribute to the formation of the ICM. However, how these processes work together remains unclear. Here we show that asymmetric segregation of the apical domain generates blastomeres with different contractilities, which triggers their sorting into inner and outer positions. Three-dimensional physical modelling of embryo morphogenesis reveals that cells internalize only when differences in surface contractility exceed a predictable threshold. We validate this prediction using biophysical measurements, and successfully redirect cell sorting within the developing blastocyst using maternal myosin (Myh9)-knockout chimaeric embryos. Finally, we find that loss of contractility causes blastomeres to show ICM-like markers, regardless of their position. In particular, contractility controls Yap subcellular localization, raising the possibility that mechanosensing occurs during blastocyst lineage specification. We conclude that contractility couples the positioning and fate specification of blastomeres. We propose that this ensures the robust self-organization of blastomeres into the blastocyst, which confers remarkable regulative capacities to mammalian embryos.

Hydrodynamic property of the cytoplasm is sufficient to mediate cytoplasmic streaming in the Caenorhabditis elegans embryo.

Cytoplasmic streaming is a type of intracellular transport widely seen in nature. Cytoplasmic streaming in Caenorhabditis elegans at the one-cell stage is bidirectional; the flow near the cortex ("cortical flow") is oriented toward the anterior, whereas the flow in the central region ("cytoplasmic flow") is oriented toward the posterior. Both cortical flow and cytoplasmic flow depend on non-muscle-myosin II (NMY-2), which primarily localizes in the cortex. The manner in which NMY-2 proteins drive cytoplasmic flow in the opposite direction from remote locations has not been fully understood. In this study, we demonstrated that the hydrodynamic properties of the cytoplasm are sufficient to mediate the forces generated by the cortical myosin to drive bidirectional streaming throughout the cytoplasm. We quantified the flow velocities of cytoplasmic streaming using particle image velocimetry (PIV) and conducted a three-dimensional hydrodynamic simulation using the moving particle semiimplicit method. Our simulation quantitatively reconstructed the quantified flow velocity distribution resolved through PIV analysis. Furthermore, our PIV analyses detected microtubule-dependent flows during the pronuclear migration stage. These flows were reproduced via hydrodynamic interactions between moving pronuclei and the cytoplasm. The agreement of flow dynamics in vivo and in simulation indicates that the hydrodynamic properties of the cytoplasm are sufficient to mediate cytoplasmic streaming in C. elegans embryos.

A Tug-of-War between Cell Shape and Polarity Controls Division Orientation to Ensure Robust Patterning in the Mouse Blastocyst.

Oriented cell division patterns tissues by modulating cell position and fate. While cell geometry, junctions, cortical tension, and polarity are known to control division orientation, relatively little is known about how these are coordinated to ensure robust patterning. Here, we systematically characterize cell division, volume, and shape changes during mouse pre-implantation development by in toto live imaging. The analysis leads us to a model in which the apical domain competes with cell shape to determine division orientation. Two key predictions of the model are verified experimentally: when outside cells of the 16-cell embryo are released from cell shape asymmetry, the axis of division is guided by the apical domain. Conversely, orientation cues from the apical domain can be overcome by applied shape asymmetry in the 8-cell embryo. We propose that such interplay between cell shape and polarity in controlling division orientation ensures robust patterning of the blastocyst and possibly other tissues.

Corrigendum: Pulsatile cell-autonomous contractility drives compaction in the mouse embryo.

This corrects the article DOI: 10.1038/ncb3185.

The Apical Domain Is Required and Sufficient for the First Lineage Segregation in the Mouse Embryo.

Mammalian development begins with segregation of the extra-embryonic trophectoderm from the embryonic lineage in the blastocyst. While cell polarity and adhesion play key roles, the decisive cue driving this lineage segregation remains elusive. Here, to study symmetry breaking, we use a reduced system in which isolated blastomeres recapitulate the first lineage segregation. We find that in the 8-cell stage embryo, the apical domain recruits a spindle pole to ensure its differential distribution upon division. Daughter cells that inherit the apical domain adopt trophectoderm fate. However, the fate of apolar daughter cells depends on whether their position within the embryo facilitates apical domain formation by Cdh1-independent cell contact. Finally, we develop methods for transplanting apical domains and show that acquisition of this domain is not only required but also sufficient for the first lineage segregation. Thus, we provide mechanistic understanding that reconciles previous models for symmetry breaking in mouse development.

Endoplasmic-reticulum-mediated microtubule alignment governs cytoplasmic streaming.

Cytoplasmic streaming refers to a collective movement of cytoplasm observed in many cell types. The mechanism of meiotic cytoplasmic streaming (MeiCS) in Caenorhabditis elegans zygotes is puzzling as the direction of the flow is not predefined by cell polarity and occasionally reverses. Here, we demonstrate that the endoplasmic reticulum (ER) network structure is required for the collective flow. Using a combination of RNAi, microscopy and image processing of C. elegans zygotes, we devise a theoretical model, which reproduces and predicts the emergence and reversal of the flow. We propose a positive-feedback mechanism, where a local flow generated along a microtubule is transmitted to neighbouring regions through the ER. This, in turn, aligns microtubules over a broader area to self-organize the collective flow. The proposed model could be applicable to various cytoplasmic streaming phenomena in the absence of predefined polarity. The increased mobility of cortical granules by MeiCS correlates with the efficient exocytosis of the granules to protect the zygotes from osmotic and mechanical stresses.

Bayesian Inference of Forces Causing Cytoplasmic Streaming in Caenorhabditis elegans Embryos and Mouse Oocytes.

Cellular structures are hydrodynamically interconnected, such that force generation in one location can move distal structures. One example of this phenomenon is cytoplasmic streaming, whereby active forces at the cell cortex induce streaming of the entire cytoplasm. However, it is not known how the spatial distribution and magnitude of these forces move distant objects within the cell. To address this issue, we developed a computational method that used cytoplasm hydrodynamics to infer the spatial distribution of shear stress at the cell cortex induced by active force generators from experimentally obtained flow field of cytoplasmic streaming. By applying this method, we determined the shear-stress distribution that quantitatively reproduces in vivo flow fields in Caenorhabditis elegans embryos and mouse oocytes during meiosis II. Shear stress in mouse oocytes were predicted to localize to a narrower cortical region than that with a high cortical flow velocity and corresponded with the localization of the cortical actin cap. The predicted patterns of pressure gradient in both species were consistent with species-specific cytoplasmic streaming functions. The shear-stress distribution inferred by our method can contribute to the characterization of active force generation driving biological streaming.

Pulsatile cell-autonomous contractility drives compaction in the mouse embryo.

Mammalian embryos initiate morphogenesis with compaction, which is essential for specifying the first lineages of the blastocyst. The 8-cell-stage mouse embryo compacts by enlarging its cell-cell contacts in a Cdh1-dependent manner. It was therefore proposed that Cdh1 adhesion molecules generate the forces driving compaction. Using micropipette aspiration to map all tensions in a developing embryo, we show that compaction is primarily driven by a twofold increase in tension at the cell-medium interface. We show that the principal force generator of compaction is the actomyosin cortex, which gives rise to pulsed contractions starting at the 8-cell stage. Remarkably, contractions emerge as periodic cortical waves when cells are disengaged from adhesive contacts. In line with this, tension mapping of mzCdh1(-/-) embryos suggests that Cdh1 acts by redirecting contractility away from cell-cell contacts. Our study provides a framework to understand early mammalian embryogenesis and original perspectives on evolutionary conserved pulsed contractions.

A cellular funicular: A hydrodynamic coupling between the anterior- and posterior-directed cytoplasmic flows.

Organelles inside cells move to position themselves at the right place at the right time. A mechanism for generating active force exists for each of such directed organelle movements. In our recent study on cytoplasmic streaming in the Caenorhabditis elegans one-cell embryo, we demonstrated that an anterior-directed force generated by myosin could drive not only anterior-directed cortical flow but also posterior-directed cytoplasmic flow. This coupling of flows in opposing directions is mediated by the hydrodynamic properties of the cytoplasm. This work provided a good example of an active force generation mechanism that drives organelle movements in two opposite directions inside the cell, just as a funicular moves up and down a slope. Interestingly, the funicular-like coupling of intracellular movements is also seen in our recent studies on centrosome positioning in the C. elegans embryo and on interkinetic nuclear movement during mouse neurogenesis. Thus, funicular-like coupling may be a general strategy used repeatedly in cells. The use of the funicular-like coupling seems advantageous because it is efficient, as one active force generation mechanism can drive movements in two directions, and also because the two movements can be coordinated to have similar speeds.