Biology subjects

Song, Z.

Publications and source records attributed to Song, Z..

At least 19 recordsLinked to original sources

bioRxiv2026

Relative reward memories retain multidimensional information in bumblebees

Previously, we showed that bumblebees encode reward value relative to local alternatives and, when comparing options learned in separate contexts, retrieve only their ordinal rank rather than absolute reward magnitude (Solvi et al., 2022). However, that study, like related work across species, varied reward along a single dimension. Natural rewards such as nectar are multidimensional, yet how such rewards are represented in relative memory remains unknown. We tested this by training bumblebees in separate contexts with rewards varying in two intrinsic dimensions, sucrose concentration and NaCl content, and then comparing options never encountered together. When only sucrose concentration varied, bees showed no preference between options that had each been locally better, despite differing in absolute concentration. With multidimensional rewards, however, this equivalence no longer held: bees preferred an option that had been locally better on both sucrose and NaCl over one better on only sucrose or only NaCl. A two-dimensional advantage also produced a stronger preference than a one-dimensional advantage 48 h after training. Thus, the ordinal rank we previously identified captures only part of the relative reward representation: with multidimensional rewards, bumblebees also retain information about the dimensions on which an option had been locally better.

animal behavior and cognition↗

bioRxiv2026

LiverHomo: An Integrated and Interactive Human Liver Single-Cell Atlas with Protein Interaction and Cell-Cell Communication Networks

LiverHomo is an integrated cell atlas developed from publicly available human liver single-cell RNA sequencing (scRNA-seq) datasets spanning healthy and diverse liver disease contexts. While several single-cell resources for the human liver are available, LiverHomo provides broad disease coverage and extends transcriptomic exploration through the integration of protein-protein interaction (PPI) networks and cell-cell interactions. LiverHomo provides an interactive web portal (https://liverhomo.org) for exploring cellular composition, gene expression, differentially expressed genes, Gene Ontology annotations, cell-cell interactions, and context-specific PPI networks across diseases and cell types. To facilitate intuitive access to these multidimensional data, the portal incorporates Ask LiverHomo, a natural-language assistant that interprets user queries and directs users to relevant analyses and visualizations. Interactive histograms, violin plots, network visualizations, and metadata exploration enable users to examine the atlas at cellular, transcriptomic, and protein-interaction levels, while the underlying data are readily accessible for further analysis. Together, LiverHomo provides an integrated and accessible resource for investigating cellular and molecular alterations across human liver health and disease.

bioinformatics↗

bioRxiv2026

An implantable sensor network for large-scale cellular force mapping in living embryos

Mechanical forces generated by cells are fundamental regulators of embryonic development, yet how locally generated forces are integrated across tissues to drive large-scale morphogenesis remains poorly understood. Addressing this question requires direct quantitative mapping of intercellular forces across tissue scale with cellular resolution while preserving native tissue architectures and dynamics, a capability not yet achieved. Here, we introduce FORCENET, an implantable force-sensing network that enables quantitative tissue-scale force mapping at cellular resolution within living embryos. FORCENET consists of a microfabricated network of ultrathin elastic sensors whose three-dimensional deformations are tracked over time and converted into intercellular force maps through physics-based reconstruction. FORCENET covers hundreds of cells while occupying a total volume 6- to 20-fold smaller than that of a single cell. Anatomical measurements, three-dimensional fluorescent imaging, and cell density and speed tracking confirmed that FORCENET seamlessly integrates within developing embryos without perturbing cell motion and proliferation, tissue architecture, and embryonic development. Applied to zebrafish gastrulation, FORCENET revealed that intercellular forces are dynamically coordinated across supracellular length scales of ~30-210 m. Inhibition of fibroblast growth factor signaling accelerated force accumulation and shortened the force-coordination length scales despite only modest changes in cell density and kinematics, demonstrating that FORCENET uncovers information not readily inferred from cell tracking alone. FORCENET establishes a scalable platform for in vivo force mapping and reveals that cell-generated forces are coordinated into supracellular mechanical patterns that are regulated by developmental signaling.

developmental biology↗

bioRxiv2026

Design of synthetic activators for programmable gene expression

Transcription factors control gene expression by recruiting transcriptional coactivators through intrinsically disordered transcriptional activation domains (TADs), whose weak and promiscuous interactions make it difficult to establish quantitative relationships between coactivator recruitment and transcriptional output, and to precisely modulate gene expression. To overcome this limitation, we set out to design coactivator-recruiting proteins using AlphaFold3-predicted TAD-coactivator interactions as a guide. We developed a high-throughput functional screening platform in mammalian cells and identified over 300 well folded nanomolar affinity binders to the transcription coactivators Mediator (multiple subunits), BRD4, NCOA, and p300; targeted interface mutations enabled systematic tuning of their binding affinities. Using a panel of p300 binders spanning nanomolar to micromolar affinities, we found an ultrasensitive relationship between coactivator-binding affinity and transcriptional output, with a sharp transcriptional activation threshold at ~500 nM. Incorporating p300 binder designs into CRISPRa yielded compact synthetic effectors that drove stronger activation than VPR across multiple genomic loci with minimal detectable cytotoxicity. Combining p300- and Mediator-specific binders further enhanced transcription, revealing functional synergy between these coactivators. Together, our work establishes a quantitative framework linking coactivator-binding affinity to transcriptional output and provides compact synthetic effectors for programmable gene activation.

synthetic biology↗

bioRxiv2026

GraPPI: A Self-Supervised Graph Encoder for Transferable Protein--Protein Interaction Modeling

Protein--Protein Interactions (PPIs) are fundamental to cellular processes and represent major targets for therapeutic development. Recent advances in machine learning for protein modeling, particularly protein language models, have enabled powerful representations of protein sequences and monomeric structures. However, compared with the increasingly generalizable representations available for individual proteins, transferable representations explicitly designed to capture protein-complex and interface contexts remain less developed. Here, we developed GraPPI, a unified graph-based deep learning framework for learning transferable representations of protein--protein complexes. At its core, PPIencoder represents protein complexes as heterogeneous residual graphs and uses self-supervised masked-edge prediction to learn interaction-aware representations that integrate sequence, three-dimensional geometry, and physicochemical properties. We then evaluated the transferability of these pretrained representations across complementary PPI tasks spanning binding-mode classification, binding-affinity prediction, and mutation-associated affinity prediction. GraPPI accurately distinguished plausible binding interfaces from perturbed or decoy binding interfaces, achieving AUROC values above 0.92 on an independent benchmark. For protein--protein binding-affinity prediction, GraPPI achieved Pearson correlation coefficients of 0.690 and 0.684 on the S90 and S79 benchmarks, respectively, with performance competitive with or exceeding that of recently reported affinity-prediction methods. GraPPI further transferred effectively to mutation-associated binding-affinity prediction, achieving a Pearson correlation of 0.84 and a mean absolute error of 1.20 kcal/mol on an independent SKEMPI benchmark. Together, these results establish GraPPI as a generalizable framework for protein-complex representation learning, extending pretrained protein representations from individual proteins toward the interaction context in which protein function is realized.

biochemistry↗

bioRxiv2026

An accumbal cholinergic ISR-HCN2 axis sustains maladaptive cue motivation

Environmental cues associated with rewarding outcomes can exert persistent control over behavior long after the reward is absent, yet the cellular adaptations that stabilize such maladaptive motivation remain unclear. Here we identify nucleus accumbens cholinergic interneurons (CINs) as causal regulators of persistent cue-driven seeking. Self-administration of addictive drugs or high-fat reward, but not sucrose, selectively increased CIN pacemaker firing. Cocaine-paired cues evoked rapid acetylcholine release in the nucleus accumbens core, and bidirectional manipulation of CINs suppressed or enhanced cue-induced cocaine seeking. Mechanistically, cocaine experience engaged the integrated stress response (ISR) in CINs, increased HCN2 expression and Ih, and stabilized a hyperexcitable state. CIN-specific ISR suppression or HCN2 knockdown reduced cocaine seeking, whereas HCN2 overexpression enhanced it. Local HCN blockade and systemic treatment with the clinically approved HCN blocker ivabradine reduced cue-driven seeking across addictive-drug and high-fat-reward paradigms while sparing sucrose seeking. These findings identify an ISR-HCN2-dependent cholinergic mechanism that stabilizes maladaptive cue-driven motivation and nominate HCN-dependent excitability as a tractable target for therapeutic intervention.

neuroscience↗

bioRxiv2026

RFOptimization: Guiding Design Optimization with All-Atom Structure Prediction

Biomolecular interactions, including protein--protein interactions, protein--nucleic acid recognition, and protein--small molecule binding, underlie a wide range of biological processes and therapeutic mechanisms. Although recent \emph{de novo} design methods can generate candidate binders for diverse molecular targets, practical design campaigns remain limited by low filter-passing rates and model-specific biases that arise when designs are optimized against a single predictor. Here, we present RFOptimization (RFO), a training-free framework for all-atom biomolecular binder optimization. RFO formulates binder improvement as a residue-wise mutational search problem, sampling candidate substitutions alternately based on gradient-guided sequence optimization using all-atom structure prediction models and a cycling-based sequence redesign strategy that alternates structure generation with an orthogonal predictor and MPNN-based sequence design to improve the \emph{in silico} success rate of RFdiffusion-generated binders within minutes of computation. To reduce overfitting to any individual structure model, candidate mutations are further evaluated with orthogonal AlphaFold3 metrics as final filters. We demonstrate the generality of RFO across diverse design settings, including classical protein binder design, ligand-binding biosensor design, cyclic peptide design, and active site-aware enzyme design.

biochemistry↗

bioRxiv2026

A family-wide atlas of human connexin docking compatibility

Gap junction (GJ) channels mediate direct intercellular communication by allowing the exchange of ions, metabolites, and signaling molecules between neighboring cells. Humans express 21 connexin (Cx) isoforms that can assemble into homotypic or heterotypic channels, creating a large potential interaction landscape that shapes tissue-specific communication networks. However, the rules governing which connexin isoforms can compatibly dock remain incompletely defined. Extracellular loop 2 (EL2) sequence features have been implicated in docking specificity and used to classify connexins into two canonical compatibility groups, K-N and H, but these assignments remain largely predictive. Most potential heterotypic connexin pairings have never been experimentally tested. This incomplete interaction map limits our ability to predict which connexin combinations can assemble, how isoform co-expression shapes intercellular communication, and how these relationships are altered or exploited in disease and engineered systems. Here, we used the FETCH (Flow Enabled Tracking of Connexosomes in HEK Cells) assay to evaluate docking compatibility across the complete human connexin family. To support family-wide compatibility mapping, we used literature-supported heterotypic interactions to define a data-driven FETCH score threshold for high-confidence interaction compatibility. Homotypic FETCH measurements varied substantially across the 21 connexin isoforms, with 15 producing mean scores above the empirical threshold. We then extended FETCH analysis to all 210 pairwise heterotypic isoform combinations. The resulting interaction landscape largely recapitulated expected motif-class relationships, including enrichment within the two canonical compatibility groups, but also identified neighboring-group interactions and unexpected cross-group pairings that represented clear exceptions to class-based predictions. Consistent with these findings, pairwise EL2 motif similarity was only modestly associated with threshold-based interaction classification, indicating that EL2 similarity alone was insufficient to predict compatibility outcomes. Together, these findings suggest that motif class provides a broad organizing framework for connexin compatibility, but that pairwise docking specificity also depends on yet-unresolved isoform-specific determinants that produce neighboring-group relationships and clear cross-group exceptions. Notably, Cx46, a lens Cx also associated with melanoma and breast cancers, emerged as a broadly permissive isoform capable of interacting with partners from both major compatibility groups and more than half of the connexin family. Together, these findings establish the first family-wide experimental atlas of human connexin docking compatibility, defining canonical interactions, previously unrecognized pairings, and exceptions to established compatibility rules. This atlas provides a foundation for defining the molecular determinants of connexin specificity, understanding how isoform diversity shapes intercellular communication, and designing gap junction channels with controlled docking behavior.

biochemistry↗

bioRxiv2026

LiverDCP: A Disease-Cell-Protein Framework for Multi-scale Modeling of Disease Biology

Understanding how molecular interactions give rise to disease phenotypes across cellular contexts remains a central challenge in biomedical research. Here, we introduce a Disease-Cell-Protein (DCP) paradigm for modeling multi-scale disease biology, which jointly represents disease states, cellular composition, and protein interaction networks within a unified graph architecture. We instantiate this paradigm in the liver as LiverDCP by integrating LiverHomo, a harmonized single-cell atlas of liver diseases, with proteome-wide predicted protein-protein interactions to construct over 280 context-specific interactomes across diverse liver disease and cellular conditions. LiverDCP employs a multi-context representation learning strategy that enables joint training across hundreds of disease-cell environments, capturing shared interaction principles while preserving context-specific variation. LiverDCP incorporates pretrained protein sequence-derived features through a geometry-aware two-phase training scheme that preserves embedding structure while improving predictive performance. The resulting DCP protein embeddings reveal extensive rewiring of protein functional states across diseases, providing a transferable representation for downstream biomedical applications. Without GWAS supervision during representation learning, LiverDCP enables disease-risk gene classification and identifies cell types through which genetic risk may act. For therapeutic target discovery, LiverDCP recovers established Phase II+ MASH targets and prioritizes previously unrecognized candidates from the unannotated proteome, with 26 of the top 50 predictions showing independent PubMed evidence related to MASH biology. Context-specific interaction analysis further provides mechanistic hypotheses for less-characterized candidates. Together, these results establish DCP as a generalizable framework for connecting molecular interactions, cellular context, genetic risk, and therapeutic opportunities across complex diseases.

systems biology↗

bioRxiv2026

Mechanisms of resilience to autosomal dominant Alzheimer's disease via oligogenic modulation of rare variants in the entorhinal cortex

Two PSEN1 E280A carriers have presented extreme protection against autosomal dominant Alzheimers disease (ADAD), with over two decades of delay for dementia onset. One of them, a male heterozygous for the RELN-COLBOS protective variant showed increased neuronal density in the entorhinal cortex 1. We conducted a deep phenotyping and genotyping study of the entorhinal cortex in protected and unprotected PSEN1 E280A cases, sporadic AD, and non-demented controls. We used single nuclei and spatial transcriptomics, whole genome sequencing, and candidate genotype-associated expression changes (GAEC) analysis. Our results showed unique neuronal and oligodendrocytic populations in the male RELN-COLBOS patient. Unique RELN positive inhibitory interneurons were enriched in cortical Layer I, while unique abundant ADAMTSL1 positive excitatory neurons were distributed in Layers II/III and Layer Va. These neurons and mature myelinated oligodendrocytes benefitted from increased expression of LRP6 receptor, functioning as a non-canonical receptor for the mutated Reelin protein encoded by RELN-COLBOS. Finally, GAEC and pathway enrichment analyses suggested that other mutations enhanced RELN-COLBOS effects in oligodendrocytes in the male RELN-COLBOS patient, explaining the phenotypic differences with his sister, a RELN-COLBOS carrier with no evident protection from ADAD. Our findings suggest that extreme deviations of the PSEN1 E280A phenotype are more likely attributed to oligogenic effects, including simultaneous mutations occurring in genes including ITGA2, involved in single molecular pathways, such as the Integrins / Focal Adhesion pathway, as potential disease modifiers for Alzheimers disease (AD).

neuroscience↗

bioRxiv2026

Mechanically-mediated low-pressure cell membrane poration enables tunable intracellular delivery of traditionally impermeable cargoes in high throughput and clinical scale formats

Traditional intracellular delivery methods suffer from cargo inefficiencies, non-linear delivery kinetics, and cellular trauma. We designed a mechanically-mediated poration platform governed by deterministic, passive diffusion operating at low pressure which enables dose-dependent, cargo-agnostic intracellular delivery and preserves cellular homeostasis; key advantages include high-fidelity multiplexing, transient cell engineering, and direct-to-biology live-cell target engagement enabling development of novel intracellular delivery applications across drug discovery and cell therapy. We demonstrate examples including live-cell DEL discovery and MOA studies, complex and rapid cell therapy manufacturing, and assay development in sensitive primary cell types.

cell biology↗

bioRxiv2026

Single-section spatial hypoxia–cytotoxic associations do not consistently reproduce across breast cancer patients

Spatial transcriptomics can reveal localized tumor-immune relationships, but thousands of spots from one tissue section do not provide thousands of biological replicates. We evaluated the distinction between within-section association and patient-level reproducibility using public breast cancer datasets. In a 10x Genomics Visium discovery section containing 3,798 spots, hypoxia-related transcription was inversely associated with cytotoxic gene activity in neighboring spots (Spearman{rho} = -0.202). High-hypoxia spots also had lower neighborhood cytotoxic scores than low-hypoxia spots (rank-biserial effect = -0.286). We then tested the directional association in an independent HER2-positive cohort comprising 36 sections, 13,619 spots, and eight patients. Only 19 of 36 sections and five of eight patients showed negative associations. The median patient-level correlation was -0.043 and did not differ from zero in a one-sided exact Wilcoxon test (P = 0.473). Sensitivity analyses using alternative cytotoxic and hypoxia signatures, neighborhood sizes, and Kendall correlation did not support a consistent inverse patient-level effect. Thus, a strong single-section association did not consistently reproduce across patients. These results caution against interpreting spot-level spatial associations from one section as patient-level biological effects.

Cancer Biology↗

bioRxiv2026

PertDiffBench: Benchmarking Diffusion Models for Single-Cell Perturbation Response Prediction

Diffusion models are increasingly used to predict transcriptional responses to perturbations, but whether they improve on simpler generative and representation-based baselines remains unclear. Existing evaluations often do not separate the effects of model architecture, input representation, biological context and metric choice, making it difficult to determine where diffusion-based methods are useful. Here we introduce PertDiffBench, a standardized benchmark for diffusion-based transcriptomic perturbation prediction across single-cell and bulk RNA-seq datasets. PertDiffBench evaluates diffusion-based models across three complementary evaluation settings: standard prediction in known single-cell contexts and bulk perturbation conditions, generalization to unseen cell types, species, drugs and intermediate time points, and stress tests of feature dimensionality, input representation, noise type and gene ordering. Across these settings, diffusion models did not show a consistent advantage. scGen remained a strong baseline in common prediction tasks, whereas scDiffusion was the most competitive diffusion-based method in several generalization settings. Temporal imputation showed a different pattern, with a simple DDPM operating directly in expression space outperforming more specialized models. Stress tests showed that performance was model dependent and sensitive to feature dimensionality, encoder choice, noise type and gene ordering. Pretrained encoders did not consistently improve performance, with the classical scVI representation slightly exceeding STATE in seen-condition and unseen-cell-type settings. These results indicate that diffusion-model performance in perturbation response prediction depends strongly on task design and representation choice. PertDiffBench provides a practical framework for evaluating these models under biologically varied and stress-tested conditions.

bioinformatics↗

bioRxiv2026

4 Phenylbutyrate Plus Gene augmentation: A dual therapy To Rescue of SLC6A1 Variant Associated Developmental And Epileptic Encephalopathy

BackgroundPathogenic variants in SLC6A1, which encodes the {gamma}-aminobutyric acid (GABA) transporter GAT-1, cause developmental and epileptic encephalopathies (DEEs) through reduced GABA uptake, impaired transporter trafficking, and functional haploinsufficiency. 4-Phenylbutyrate (PBA) is a clinically available small molecule with chemical-chaperone and histone-deacetylase-inhibitor activities that can rescue misfolded GABAergic proteins, but variant-level rescue data are needed to guide precision treatment. MethodsWe report a newly identified de novo SLC6A1 missense variant, p.Ala305Val (A305V), in a patient with myoclonic-atonic epilepsy and a developmental and epileptic encephalopathy phenotype. A305V was compared with the residue-matched comparator p.Ala305Thr (A305T). Variant effects were evaluated by (i) protein-structure prediction across nine stability-prediction algorithms using the cryo-EM-derived human GAT-1 template (PDB 7Y7W); (ii) 3H-GABA uptake assays in HEK293T cells and in human iPSC-derived astrocytes and cortical neurons; (iii) live-cell confocal microscopy of ER colocalization; (iv) pharmacologic rescue with PBA, TUDCA and salubrinal (v) and GAT-1 cDNA gene-augmentation, alone and in combination with PBA. ResultsAI-based stability predictors uniformly indicated destabilization of GAT-1(A305V) and GAT-1(A305T). A305V reduced 3H-GABA uptake across HEK293T, astrocyte, and neurons. The mutant transporter accumulated within the endoplasmic reticulum (ER), with ER colocalization rising from approximately 30% in wildtype to [~]80% in A305V; PBA reduced ER retention to approximately [~]40% and restored total GAT-1 fluorescence toward wildtype levels. Pharmacochaperones (PBA, TUDCA) restored GABA uptake for the mutant transporters. Wildtype GAT-1 gene augmentation improved mutant GAT-1 uptake and combined PBA-plus-augmentation produced rescue greater than either intervention alone in the available dose-response ranges. ConclusionsSLC6A1 A305V is a trafficking-impaired, loss-of-function GAT-1 variant whose dysfunction is tractable to two convergent therapeutic axes: pharmacologic correction of folding and trafficking, and augmentation of functional transporter dose. These findings support a two-pronged precision-medicine framework for SLC6A1-related DEEs in which PBA increased the transporter function augmented by increased gene therapy. Significance of the studyThis work links the patient-derived SLC6A1 A305V variant to a defined molecular mechanism--GAT-1 destabilization, ER retention, and reduced GABA uptake--and demonstrates that the deficit is reversible by two independent interventions that converge at the same downstream endpoint of functional surface transporter. Because PBA is already clinically deployable and GAT-1 cDNA augmentation models a future viral or non-viral gene therapy, the combined-rescue logic provides a falsifiable path for precision medicine in SLC6A1-related DEEs: chemical chaperoning corrects the folding bottleneck while transporter augmentation increases the pool available for rescue.

neuroscience↗

bioRxiv2026

Host genetic regulation of fermentation-related cecal microbial taxa is associated with loin muscle deposition in meat rabbits

Loin muscle weight is an important indicator of carcass yield in meat rabbit production, but the host genetic and intestinal microbial factors associated with its variation remain poorly understood. Because the cecum is the primary site of hindgut fermentation in rabbits, we integrated whole-genome resequencing, cecal transcriptome profiling, and cecal and rectal 16S rRNA sequencing data from 321 Kangda meat rabbits, with rectal microbiome data used as a downstream comparative reference. Compared with the rectum, the cecum contained a richer microbial community, with more ASVs and genera and significantly higher microbial diversity, whereas predicted metabolic functions were largely conserved between the two segments. Loin muscle weight showed moderate SNP-based heritability (h{superscript 2} = 0.39), and the cecal microbiome explained a smaller but detectable proportion of phenotypic variation (m{superscript 2} = 0.13). Multi-strategy microbial screening identified 19 candidate cecal genera associated with loin muscle weight, with Methanosphaera showing the strongest negative association. Host GWAS prioritized candidate loci near MEX3C and TCF4 on chromosome 10, and integration of cecal cis-eQTL and GWAS summary statistics further prioritized GJB3 as a candidate gene associated with loin muscle weight. Consistent with the cecum-centered model, host genetic relatedness was weakly but significantly correlated with cecal microbial similarity, whereas no such global association was observed for the rectal microbiome. Microbial GWAS and SMR analyses further prioritized a cecal MRAP2-Methanobrevibacter association as the main host-regulated microbial signal, while rectal analyses identified distinct segment-specific signals, including SULF1-Roseburia. These findings suggest that host genetic variation may be linked to loin muscle deposition partly through cecal gene expression and fermentation-related cecal microbial taxa, with the rectal microbiome providing comparative evidence for hindgut segment specificity. This study provides candidate host and microbial targets for future functional validation and microbiome-informed nutritional strategies to improve carcass traits in meat rabbits.

genetics↗

bioRxiv2026

Gigabase-scale deletion scanning of the human genome

What fraction of the human genome is essential for cellular viability? To date, essentiality in human cells has been mapped almost exclusively at the level of individual open reading frames (ORFs). Whether noncoding regions and broader architectural features of the genome are required for human cells to remain viable, and where the boundaries of any such regions lie, remains largely unexplored. Here we introduce Shred-seq, which couples Type I-C CRISPR-Cas3-mediated deletions with phage polymerase-based genotyping to enable large-scale deletion scans of the human genome. Shred-seq leverages thousands of genomically integrated, mapped target sites as launchpads for Cas3 to initiate unidirectional deletions ranging in size from hundreds of base pairs (bp) to hundreds of kb. Breakpoints are directly captured at high resolution by in vitro or in situ transcription from flanking phage polymerase promoters, enabling bulk or single-cell phenotyping, respectively. In this proof-of-concept, we generate and genotype 36,257 independent deletions originating from 9,604 Cas3 launchpads, individually spanning 100 bp to 500 kb, collectively covering 461 Mb (14% of the human genome), and totaling to 2.55 Gb of deleted sequence ([~]10-fold coverage of these regions). Surviving deletions are depleted not only for essential protein-coding genes, but also for active, conserved and mutation-constrained non-coding sequences, directly quantifying purifying selection across both coding and noncoding intervals. Deletion length distributions further enable annotation-agnostic fine mapping of essential region boundaries and establish an empirical lower bound on the fraction of the human genome required for cellular viability. Finally, we demonstrate compatibility with single-cell RNA-seq (scRNA-seq), enabling direct linkage of specific deletions to transcriptional phenotypes. Together, Shred-seq provides a scalable platform for the systematic dissection of genome architecture and noncoding function, as well as for generating training data for predictive and generative models of genome structure-function, analogous to the roles that conventional Perturb-seq screens play for ORFs. By enabling the empirical delimitation of the genomic content required for human cellular viability, Shred-seq may also open a path to the construction of a minimal human genome.

genomics↗

bioRxiv2026

4-Phenylbutyrate Rescue in GABRA1 Variants Associated with Developmental Epileptic Encephalopathies: from Cell and Mouse Models to Human

Disease variants in GABR genes encoding {gamma}-Aminobutyric acid type A receptor (GABAAR) subunits are major causes of developmental and epileptic encephalopathies (DEEs). There is no effective treatment for these DEEs although the GABAAR is a major target for antiseizure drugs. We previously identified the therapeutic effect of 4-phenyl-butyrate (PBA) in Gabrg2+/Q390X knockin DEE mice and now test the effect of the drug in GABRA1 variants that encode the 1 subunit. We used a multidisciplinary approach including in silico structural modeling, flow cytometry, patch clamp recordings and bio-chemistry in conjunction with differential tagging of the wild-type and the mutant alleles to evaluate the effect of PBA on rescue of GABAAR subunit expression, surface trafficking, and function in vitro in heterologous HEK293T cell model and in vivo in Gabra1+/A322D mice. We found that both total and cell surface 1 expression was reduced when the variant 1 protein was present; suggesting reduced functional receptor on the cell membrane and synapse. Patch clamp recordings identified 1 variants reduced GABA-evoked current amplitude. In silico prediction indicated reduced protein stability for GABRA1 variants indicated by negative {Delta}{Delta}G values. PBA increased both total and surface expression of wildtype 1 and 1 variants; and improved expression of both wildtype and variant 1 alleles when these were co-expressed. Importantly, PBA also increased the GABAAR expression in the thalamus of the Gabra1+/A322D mice. This study indicates that PBA is a promising treatment option for DEEs associated with GABRA1 mutations. Our previous work has demonstrated that PBA improves proteostasis by enhancing expression of the wildtype allele, repairing the mutant allele, and reducing endoplasmic reticulum stress. Therefore, it can mitigate seizures and improve neurobehavioral phenotypes at behavioral levels. Based on this and our previous work on GABRG2 and SLC6A1 mutations, we propose that PBA holds promise as a common medicine for multiple genetic neurologic disorders that share the proteostasis pathology with a broad clinical application in DEEs.

neuroscience↗

bioRxiv2026

Zero-Shot Generation of Protein Conformational Ensembles Through AlphaFold Latent Flooding

Despite transformative advances in protein structure prediction, generating conformational ensembles directly from sequence in an efficient and accessible manner remains a central challenge. We introduce a heuristic importance sampling framework for the zero-shot prediction of functionally relevant, conformationally diverse structures, bypassing the need for extensive physics-based modeling or prior domain knowledge. Guided by an analysis of massive activations in AlphaFold (AF), we develop a latent flooding algorithm that enables adversarial exploration of the latent space while preserving local geometric integrity. We show that the AF latent flooding (AFLF) method robustly recovers experimental structural fluctuations, captures function-relevant conformational states, and reveals cryptic binding sites across diverse protein systems. These results suggest that AF latent features implicitly encode the biophysical principles of protein thermodynamics, which AFLF exploits. Finally, we show that AFLF is computationally accessible and interoperable, offering a test-time generalization of AF for investigating protein dynamics and accelerating structure-based ligand discovery.

biophysics↗