Jaspreet Singh, Ph.D.Flatiron Research Fellow, Developmental Dynamics, CCB, Flatiron Institute
Zijun Zhang, Ph.D.Assistant Professor, Division of Artificial Intelligence in medicine at Cedars-Sinai Medical Center
1st Speaker: Jaspreet Singh, Ph.D., Flatiron Research Fellow, Developmental Dynamics
Title: Biophysics of Arrested Furrows
Absrtact: Cell division is often incomplete in early-stage gametogenesis: instead of separating, daughter cells remain connected through cleavage furrows that are stabilized into intercellular bridges (ICBs). While significant progress has been made in elucidating their ultrastructural and molecular composition, the dynamics of their assembly and mechanical stability is poorly understood. We develop a one kinematic variable model that accounts for geometrical nonlinearities, underlying the dynamics of transition from a parent cell to two daughter cells with an ICB. We show that precise control of contractility and furrow remodeling is essential for assembling ICBs with experimentally observed aspect ratios.
2nd Speaker: Frank Zijun Zhang, Ph.D., Flatiron Research Fellow, Genomics
Title: AMBER: An automated framework for efficiently designing deep convolutional neural networks in genomics
Abstract: Convolutional neural networks (CNNs) have become a standard for analysis of biological sequences. Tuning of network architectures is essential for CNN’s performance, yet it requires substantial knowledge of machine learning and commitment of time and effort. AMBER is a fully automated framework to efficiently design and apply CNNs for genomic sequences. We applied AMBER to the tasks of modelling genomic regulatory features and CRISPR/Cas9 editing outcomes, demonstrating that the predictions of the AMBER-designed model outperformed the equivalent baseline non-NAS models, even the published expert-designed models. Interpretation of AMBER architecture search revealed its design principles of utilizing the full space of computational operations for accurately modelling genomic sequences.