1st Presenter: Reza Farhadifar, Ph.D., Research Scientist, Biophysical Modeling
Topic: Tissue fluidity regulates ommatidial rotation in Drosophila eye
Abstract: The phenomenon of tissue fluidity, the ability of cells to rearrange relative to each other in a confluent tissue, has been linked to several morphogenetic processes and diseases, yet the molecular regulators of tissue fluidity remain unknown. Ommatidial rotation is a morphogenetic process in the Drosophila eye driven by planar cell polarity signaling and shares many features with polarized cellular migration in vertebrates. Rotation requires the concerted regulation of force transmission between cells via adherens junctions, but how other cellular mechanisms coordinate motility and tissue fluidity remains poorly understood. We developed in vivo live imaging and analysis tools to quantify cellular morphologies and dynamics, and these revealed that rotation is driven autonomously by ommatidial cells that cluster together and rotate in successive pulses within the permissive substrate of interommatidial cells. By analyzing a rotation-specific nemo mutant (Nlk in vertebrates), we demonstrate that precise regulation of junctional E-cadherin levels is critical for modulating the tissue’s mechanical properties, allowing rotation to progress. Nemo promotes the turnover of E-cadherin, modulating adherens junction tension and overall epithelial fluidity. Our study defines the Nemo kinase as a molecular regulator to induce a transition from a solid-like tissue with high internal stress to a more viscoelastic tissue, and broadens our molecular understanding of tissue fluidity.
2nd Presenter: Vladimir Gligorijevic, Ph.D., Research Scientist, Systems Biology
Topic: Function-guided protein design by manifold sampling
Abstract: The protein design space arising from all possible combinations of amino acids is a large combinatorial space that is only sparsely functional. Traditional computational methods sample new sequences via computationally expensive simulations. Thus, efficiently exploring the vast protein sequence space still remains a challenge.
I will introduce a new method that combines ideas from self-supervised representation learning, active learning and generative probabilistic modeling to learn the protein sequence space of all existing proteins and navigate through it more efficiently. Our method is tightly integrated with state-of-the-art protein modeling and design platforms and offers promising new directions for function-focused protein design.