Biology subjects

Takeshita, S. S.

Publications and source records attributed to Takeshita, S. S..

3 recordsLinked to original sources

bioRxiv2026

Transforming macromolecular structures into simulations of self-assembly

Macromolecular self-assembly is a fundamental process in living and engineered systems, producing molecular machines like the ribosome or highly symmetric viral capsids. Thanks to sources like the Protein Data Bank (PDB) and AlphaFold3, the final target complexes are often known, but these static structures do not provide information on the self-assembly process directly. Computational models provide critical tools to study these essential pathways of self-assembly, but substantial coarse-graining of assembly subunits is necessary to achieve computational tractability of these relatively slow processes while retaining multi-valency. While rule-based or local interactions overcome the often-prohibitive enumeration of all possible assembly intermediates, they must ensure global structural constraints are met. We here demonstrate ioNERDSS, a user-friendly Python package that transforms 3D atomic structures into coarse-grained models for immediate simulation with the stochastic reaction-diffusion NERDSS software, converting static structures into time-resolved assembly trajectories. NERDSS uses rule-based interactions to simulate multi-component self-assembly at the minutes timescales and without limits to complex size or growth pathways. With ioNERDSS, each protein chain is defined by a rigid subunit with discrete interfaces and explicit orientational constraints that enforce a structured assembly. Repeated subunits (such as in viral capsids) are regularized to preserve the target topology across distinct stochastic assembly pathways, supporting assembly of structures with thousands of subunits. We initialize pairwise binding affinities using open-source machine-learned prediction tools, and our default coarse-grained (CG) models are all constrained by thermodynamic reversibility to reach an equilibrium steady-state. The binding rates and subunit abundances necessary to perform simulations are initialized at default values but represent the key variables (along with affinities) that cells and thus users would tune to control productive assembly. Benchmarking on over 40,000 PDB structures shows that the majority of CG models stochastically assemble into target structures. The ioNERDSS Python library links directly to open-source tools for visualization and analysis to facilitate fast and user-friendly structure validation and analysis of output for thermodynamic, kinetic, and nonequilibrium drivers of macromolecular self-assembly.

biophysics↗

bioRxiv2025

A membrane-driven biochemical oscillator tunable by the volume to surface area ratio

Oscillations are ubiquitous features of biological organisms, playing crucial roles in processes from circadian rhythms to developmental patterning. Protein-based biochemical oscillators have particular applications in synthetic biology because they can access fast and slow timescales that are independent from the transcription-translation machinery required of genetic oscillators. Here, we introduce and model such a mass-conserving biochemical oscillator using mass-action reaction kinetics that exploits dynamic changes to membrane phospholipid concentrations to drive proteins on and off the membrane in robust, tunable rhythms. Importantly, the oscillations rely on amplification of reactions on the membrane via dimensional reduction, and they are therefore tunable by variations in the volume-to-surface area ratio (V/A) of the system. With components inspired by the endocytic machinery, we show that a wide range of physiologically relevant biochemical rates can produce oscillations in part due to this independent geometric control. A broad computational screen of the high-dimensional parameter space reveals that oscillations require relatively strict enzyme kinetic design rules for low V/A but much more permissive kinetics for larger V/A. We validate that oscillations persist with more realistic reaction-diffusion simulations that captures explicit diffusion and stochastic, integer valued copy numbers, in overall good agreement with the period and amplitude of the deterministic oscillators. Because the oscillations rely on time-dependent changes to the surface properties and not post-translational modifications to the protein subunits, we demonstrate that it can be coupled to a self-assembling trimer, driving not only changes in localization but trimer yield. Our analysis establishes this membrane-localization oscillator as a new, geometry tunable and programmable timing module and suggests a potential for geometry sensing in engineered or cell-free systems.

biophysics↗

bioRxiv2025

Evaluation of De Novo Deep Learning Models on the Protein-Sugar Interactome

Advances in deep learning have produced a range of models for predicting the protein-sugar interactome; however, structural docking of noncovalent protein-carbohydrate complexes remains largely unexplored. Although all-atom structure prediction models like AlphaFold3 (AF3), Boltz-1, Chai-1, DiffDock, and RosettaFold-All Atom (RFAA) were validated on protein-small molecule complexes, no benchmark or evaluation exists specifically for noncovalent protein-carbohydrate docking. To address this, we developed a high-quality dataset of experimental structures - Benchmark of CArbohydrate Protein Interactions (BCAPIN). Using BCAPIN and a novel evaluation metric, DockQC, we assessed the performance of all-atom structure prediction models on non-covalent protein-carbohydrate docking. We found all methods achieved comparable results, with an 85% success rate for structures of at least acceptable quality. However, we found that the predictive power of all models declined with increasing carbohydrate polymer length. With the capabilities and limitations assessed, we evaluated AF3s ability to predict binding for a set of putative human carbohydrate binding and carbohydrate non-binding proteins. While current models show promise, further development is needed to enable high-confidence, high-throughput prediction of the complete protein-sugar interactome.

biophysics↗