Search for: All records

Abstract We introduce the kernel-elastic autoencoder (KAE), a self-supervised generative model based on the transformer architecture with enhanced performance for molecular design. KAE employs two innovative loss functions: modified maximum mean discrepancy (m-MMD) and weighted reconstruction (LWCEL). The m-MMD loss has significantly improved the generative performance of KAE when compared to using the traditional Kullback–Leibler loss of VAE, or standard maximum mean discrepancy. Including the weighted reconstruction loss LWCEL, KAE achieves valid generation and accurate reconstruction at the same time, allowing for generative behavior that is intermediate between VAE and autoencoder not available in existing generative approaches. Further advancements in KAE include its integration with conditional generation, setting a new state-of-the-art benchmark in constrained optimizations. Moreover, KAE has demonstrated its capability to generate molecules with favorable binding affinities in docking applications, as evidenced by AutoDock Vina and Glide scores, outperforming all existing candidates from the training dataset. Beyond molecular design, KAE holds promise to solve problems by generation across a broad spectrum of applications. 

Abstract Methane clathrates on continental margins contain the largest stores of hydrocarbons on Earth, yet the role of biomolecules in clathrate formation and stability remains almost completely unknown. Here, we report new methane clathrate-binding proteins (CbpAs) of bacterial origin discovered in metagenomes from gas clathrate-bearing ocean sediments. CbpAs show similar suppression of methane clathrate growth as the commercial gas clathrate inhibitor polyvinylpyrrolidone and inhibit clathrate growth at lower concentrations than antifreeze proteins (AFPs) previously tested. Unlike AFPs, CbpAs are selective for clathrate over ice. CbpA3 adopts a nonglobular, extended structure with an exposed hydrophobic surface, and, unexpectedly, its TxxxAxxxAxx motif common to AFPs is buried and not involved in clathrate binding. Instead, simulations and mutagenesis suggest a bipartite interaction of CbpAs with methane clathrate, with the pyrrolidine ring of a highly conserved proline residue mediating binding by filling empty clathrate cages. The discovery that CbpAs exert such potent control on methane clathrate properties implies that biomolecules from native sediment bacteria may be important for clathrate stability and habitability. 

The cytoskeleton is a major focus of physical studies to understand organization inside cells given its primary role in cell motility, cell division, and cell mechanics. Recently, protein condensation has been shown to be another major intracellular organizational strategy. Here, we report that the microtubule crosslinking proteins, MAP65-1 and PRC1, can form phase separated condensates at physiological salt and temperature without additional crowding agents in vitro. The size of the droplets depends on the concentration of protein. MAP65 condensates are liquid at first and can gelate over time. We show that these condensates can nucleate and grow microtubule bundles that form asters, regardless of the viscoelasticity of the condensate. The droplet size directly controls the number of projections in the microtubule asters, demonstrating that the MAP65 concentration can control the organization of microtubules. When gel-like droplets nucleate and grow asters from a shell of tubulin at the surface, the microtubules are able to re-fluidize the MAP65 condensate, returning the MAP65 molecules to solution. This work implies that there is an interplay between condensate formation from microtubule-associated proteins, microtubule organization, and condensate dissolution that could be important for the dynamics of intracellular organization.