Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/2024.08.04.606514

FoldToken4: Consistent & Hierarchical Fold Language

Gao, Z.·Tan, C.·Li, S. Z.
View original source ↗Export metadata ↓

Abstract

Creating protein structure language has attracted increasing attention in unifing the modality of protein sequence and structure. While recent works, such as FoldToken1&2&3 have made great progress in this direction, the relationship between languages created by different models at different scales is still unclear. Moreover, models at multiple scales (different code space size, like 25, 26, [···], 212) need to be trained separately, leading to redundant efforts. We raise the question: Could a single model create multiscale fold languages? In this paper, we propose FoldToken4 to learn the consistent and hierarchical of multiscale fold languages. By introducing multiscale code adapters and token mixing techniques, FoldToken4 can generate multiscale languages from the same model, and discover the hierarchical token-mapping relationships across scales. To the best of our knowledge, FoldToken4 is the first effort to learn multi-scale token consistency and hierarchy in VQ research; Also, it should be more novel in protein structure language learning.

Explore related subjects

molecular biologyPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Gao, Z., Tan, C., Li, S. Z.. 2024-08-04. FoldToken4: Consistent & Hierarchical Fold Language. https://doi.org/10.1101/2024.08.04.606514

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Trans-branching of polyubiquitin chains orchestrates the DNA replication stress response

Polyubiquitin chain geometry dictates functional consequences of ubiquitylation. Although branched polyubiquitin chains are abundant in cells, little is known about their functions. Here we show that branching on the DNA replication factor PCNA, mediated by the ubiquitin-conjugating enzyme UBE2K and involving lysines 63 and 48 of ubiquitin, orchestrates the sequence of events in response to replication stress. By inducing VCP-dependent extraction of PCNA from chromatin, branching promotes re-priming of stalled forks and necessitates a BRCA1-dependent pathway of daughter-strand gap repair. Our study identifies hyper-accumulation of daughter-strand gaps as the mechanistic basis underlying the toxicity of inhibitors of the PCNA-specific isopeptidase, USP1, in BRCA1-deficient cells. Moreover, an unexpected preference of UBE2K to operate in trans suggests a general timing mechanism to organize hierarchies amongst ubiquitin signals.

molecular biology↗
bioRxiv2026

Impaired proteostasis is an early feature of the diabetic heart in humans and mice

Diabetes and obesity increase cardiac lipid levels leading to cardiomyopathy and heart failure. We hypothesized that intermittent fasting would reduce cardiac lipid levels. Surprisingly, intermittent fasting increased myocardial triglyceride content, but rescued mortality and attenuated cardiomyopathy in mice overexpressing cardiomyocyte acyl-CoA synthetase 1 (MHC-ACSL1). Lipid overload caused cardiomyocyte accumulation of polyubiquitinated protein aggregates containing desmin, a scaffolding intermediate filament protein, which intermittent fasting prevented. Furthermore, intermittent fasting reversed elevated myocardial C16:0 ceramide content, and knockdown of ceramide synthase CerS5 and CerS6 reduced palmitate-induced protein aggregation, highlighting a role for C16:0 ceramides in this pathology. Conversely, impairing aggrephagy with cardiomyocyte-specific p62 ablation induced heart failure in mice fed a high-fat diet, with paradoxically reduced cardiac lipid content. Crucially, non-failing diabetic human hearts also exhibited protein aggregate pathology. Taken together, these results demonstrate that impaired proteostasis characterizes cardiomyopathy from cardiac lipid overload and identify a promising new therapeutic target for this condition.

molecular biology↗
bioRxiv2026

Spatial profiling and neurovascular communication in the developing and adolescent cortex following prenatal alcohol exposure

Fetal alcohol spectrum disorders (FASD) constitute a wide range of developmental, cognitive, and behavioral impairments caused by prenatal alcohol exposure (PAE). Although neuronal and vascular consequences of PAE have been studied, how alcohol affects the cerebrovasculature within the framework of the neurovascular unit (NVU) across development remains poorly understood. At minimum, the NVU comprises neurons, astrocyte endfeet, and endothelial cells (ECs), which coordinate to maintain brain homeostasis. Here, we used the NanoString Digital Spatial Profiling platform to characterize spatial transcriptomic data from neurons, astrocytes, and ECs from PAE and saccharin (SAC) control cortices at embryonic day 18 (E18) and postnatal day 28 (P28). Differentially expressed genes were then used for Ingenuity Pathway Analysis (IPA) to identify altered biological pathways and perform comparison analyses across developmental time points, while CellChat was used to infer cell cell communication networks. We uncovered thousands of differentially expressed genes and numerous altered pathways and biological processes in PAE cortices across development. Both IPA and CellChat analyses implicated dysregulation of vascular and extracellular matrix (ECM) remodeling, cell adhesion, and neuroinflammatory signaling. CellChat further predicted the loss of several key bidirectional relationships and altered ligand-receptor interactions among neurovascular cell types at E18 and P28. Overall, these findings identify PAE associated alterations in neurovascular gene expression and intercellular signaling across development, providing potential mechanisms by which PAE may disrupt neurodevelopment.

molecular biology↗