Biology subjects

Liu, X.

Publications and source records attributed to Liu, X..

At least 325 records · Page 18Linked to original sources

bioRxiv2025

Dynamic optimization of chemo-immunotherapy sequencing reveals phenotype-dependent regimens across tumor immune microenvironments

Cytotoxic chemotherapy and immune checkpoint inhibitors (ICIs) have transformed the management of advanced cancers, yet durable responses remain restricted to subsets of patients and strongly depend on the tumor immune microenvironment (TIME). Distinct "hot" and "cold" TIMEs differ in pre-existing effector T-cell activity and treatment-induced immunogenicity, suggesting that combination regimens should be tailored to microenvironmental context under realistic clinical constraints. Here we develop a deterministic dynamic optimization framework that jointly designs chemotherapy and ICI dosing schedules on a mechanistic tumor-immune model. The model tracks sensitive and resistant tumor cells, effector CD8+ T cells and circulating drug concentrations with pharmacodynamic couplings encoding immunogenic cell death and checkpoint inhibition. We formulate a continuous-time optimal control problem that balances tumor burden against drug exposure while enforcing cumulative dose limits, end-of-interval concentration bounds and bounds on stepwise changes in infusion rates. Using gradient-based optimization with embedded forward sensitivities, we solve the resulting nonlinear programs for four representative TIME phenotypes (extremely cold, hot, cold and cold with high antigenic resistance). The optimizer recovers qualitatively distinct and biologically interpretable regimens, including chemotherapy-dominated schedules in extremely cold TIMEs, aggressive early ICI administration in hot TIMEs and minimal ICI pulses in cold TIMEs. Alternative weightings of treatment objectives reveal trade-offs between tumor eradication and effector preservation. These results show that constrained dynamic optimization can systematically derive phenotype-specific, mechanistically consistent combination strategies and provide a reusable computational module for integrating TIME-resolved models into in silico trial and quantitative systems pharmacology workflows. Author summaryImmune checkpoint inhibitors (ICI) have transformed cancer therapy, yet their benefit varies dramatically across tumor immune microenvironments (TIMEs). In the clinic, ICI is often combined with chemotherapy, and clinicians must decide not only whether to use ICI, but also how to schedule both agents over months of treatment. We develop a reusable dynamic optimization framework that takes any ordinary differential equation model of tumor-immune-therapy interactions as input, enforces clinically motivated dosing and toxicity constraints, and outputs optimal chemo-ICI regimens. Using a mechanistic model and four virtual patients representing hot, cold, and extremely cold TIMEs, we show that the framework automatically suppresses futile ICI dosing in immune deserts, while allocating aggressive combination therapy to highly ICI-responsive tumors.

bioinformatics↗

bioRxiv2025

COMPARATIVE ANALYSIS OF TWO-DIMENSIONAL AND THREE-DIMENSIONAL MRI RADIOMICS MODELS FOR PREDICTING MOLECULAR BIOMAKERS IN BREAST CANCER.

BackgroundBreast cancer (BC) incidence continues to rise globally, with diverse clinical presentations that remain challenging to predict. PurposeTo compare the predictive performance of conventional 2D DCE-MRI-based radiomics models with advanced 3D DCE-MRI-based radiomics models for determining estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER-2), and Ki-67 expression in breast cancer patients. Material and MethodsThis study included 385 patients with histologically confirmed breast cancer. Semi-automatic segmentation was applied to delineate regions of interest (ROIs) on DCE-MRI images, and radiomic features were extracted using Intelligence Foundry (IF) 3.1. Feature selection involved single-value filtering, correlation analysis, outlier optimization, normalization, and least absolute shrinkage and selection operator (LASSO) regression. Logistic regression was used for statistical analysis, and model performance was evaluated using area under the receiver operating characteristic curve (AUC), decision curve analysis (DCA), and DeLongs test for comparing 2D and 3D models. ResultsThe 3D models consistently outperformed 2D models across all biomarkers. In the validation set, the 3D models achieved AUCs of 0.884 for ER, 0.861 for PR, 0.898 for HER-2, and 0.886 for Ki-67, while the 2D models yielded AUCs of 0.813 (ER), 0.740 (PR), 0.821 (HER-2), and 0.796 (Ki-67). DCA demonstrated greater net clinical benefit for 3D models, and DeLongs test revealed statistically significant differences between 2D and 3D models (p < 0.05) for all biomarker groups. DiscussionBoth 2D and 3D DCE-MRI radiomics models demonstrated substantial potential for predicting molecular marker expression in breast cancer. However, 3D models exhibited superior performance, likely attributable to their comprehensive capture of spatial tumor heterogeneity. These findings support the preferential use of 3D radiomics models in future research and clinical practice for non-invasive biomarker assessment.

cancer biology↗

bioRxiv2025

Ferrapoptosis: Discovery of a Hybrid Programmed Cell Death

Traditional taxonomies classify ferroptosis and apoptosis as distinct forms of regulated cell death. Here, we challenge this view by defining and validating a previously unrecognized neuronal death entity in traumatic brain injury (TBI), which we term "Ferrapoptosis". This hybrid mode of death is characterized by the coexistence of ferroptotic and apoptotic molecular and ultrastructural features within the same neuron-- ultrastructural alterations that cannot be fully characterized by any known cell death modalities. By integrating single-cell transcriptomics with multi-layered functional assays, we show that Ferrapoptosis dominates in severe injury and the acute phase, and is critically driven by mitochondrial oxidative stress. Over time, its predominance is gradually replaced by death modes in which either ferroptosis or apoptosis alone becomes the major pathway. Genome-wide CRISPR screening further identifies Smg7 as a key regulator that synchronously activates both death programs to drive the hybrid phenotype, whereas genetic or viral inhibition of Smg7 reduces Ferrapoptosis and promotes neurological recovery after TBI in mice. Our work systematically delineates a previously unrecognized form of cell death and its pathogenetic mechanisms, providing experimental evidence to refine cell death classification, and suggesting a conceptual strategy for treating complex diseases such as TBI by targeting shared regulatory nodes. HIGHLIGHTSO_LIDefined and validated a previously unrecognized cell death modality, termed "Ferrapoptosis". C_LIO_LIElucidated the ordered pattern of Ferrapoptosis across key stages of TBI pathogenesis. C_LIO_LIRevealed mitochondrial oxidative stress as the pivotal hub integrating ferroptosis and apoptosis. C_LIO_LIIdentify Smg7 as a core driver of Ferrapoptosis and a promising neuroprotective target. C_LI

cell biology↗

bioRxiv2025

Spatial patterning of transcriptional and regulatory programs in the primate subcortex

Mammalian brain cell identity is shaped by intrinsic factors and external context. We present a spatially resolved transcriptomic and gene regulatory atlas of cell types across all subcortical regions in a primate, the common marmoset. Dense sampling and cross-species integration revealed spatially precise neuronal assemblies, including in complex midbrain and diencephalic structures. Chromatin accessibility and transcriptional identity are spatially tuned within and across subcortical structures; spatial gradients within hippocampal subfields are orchestrated by graded transcription factors acting through graded enhancers. The primate-expanded thalamic GABAergic population shares transcriptional and regulatory syntax with conserved midbrain populations, reflecting an evolutionary adaptation compared with rodents. Similar regional expression across cell types can arise by distinct regulatory architectures, as for telencephalic astrocytes and neurons. Conversely, distant cell types can share regulatory programs despite divergent identities: striatal GABAergic medium spiny neurons and telencephalic glutamatergic neurons share a postsynaptic regulatory program despite divergent lineage, region, and neurotransmitter identity.

neuroscience↗

bioRxiv2025

The Human-In-the-Loop Drug Design Framework with Equivariant Rectified Flow

Advancements in AI-based drug design often face obstacles due to incomplete datasets, hindering progress to clinical trials. Human experts bring invaluable expertise and nuanced contextual understanding to drug design. There are two main difficulties in integrating human knowledge into the drug development process. First, human annotations are costly, and traditional machine learning algorithms require a large number of samples to be effective. Second, human experts are unable to accurately describe their expertise using natural language, nor can they precisely specify what kind of molecule is needed before seeing the generated molecules. To address these problems, we propose a new platform, called HIL-DD, for experts to infuse their experience by selecting molecules generated by AI that meet their criteria, or discarding those that do not. The core generative technology utilizes an Equivariant Rectified Flow Model (ERFM), which offers faster generation speeds than conventional diffusion models, enabling efficient human-AI collaboration. More importantly, we provide a user-friendly interface to ensure smooth and effective collaboration between human experts and AI systems. Rigorous experiments demonstrate that our system can produce 3D molecules that align with expert expectations in minimal interactive sessions. These generated molecules maintain drug-like qualities comparable to those created by current state-of-the-art models.

bioinformatics↗

bioRxiv2025

Single-Cell Multi-Omics Dissects Transcript Isoform and Immune Repertoire Dynamics in Human Immunosenescence

Immunosenescence, a major hallmark of systemic aging, refers to the progressive functional decline of the immune system. This decline not only compromises host defense and immunological memory but also fuels chronic inflammation and tissue degeneration (collectively known as inflammaging). While single-cell RNA sequencing (scRNA-seq) has revealed transcriptomic alterations in immune aging, analyses restricted to transcript abundance fail to capture deeper regulatory layers, such as transcript isoform diversity and the remodeling of immune receptor repertoires. To address this, we present the human peripheral immune single-cell multi-omics atlas that integrates gene expression, transcript isoforms diversity, and immune receptor repertoires. By combining single-cell full-length transcriptome sequencing (scCycloneSEQ), short-read scRNA-seq, and single-cell immune receptor sequencing (scTCR/BCR-seq), we systematically profiled peripheral blood mononuclear cells (PBMCs) from healthy young and elderly donors. Our analyses uncovered extensive age-related remodeling of immune cell composition, functional states, and TCR/BCR diversity. Notably, we identified in CD4 effector memory T cells exhibited widespread differential isoform usage (DIU), 3'UTR length variation, and a marked reshaping of cytotoxic T lymphocyte (CTL) clonotypes--all closely associated with aging-related inflammation and cellular senescence. This multi-omics atlas delineates key molecular features of immunosenescence and provides a high-resolution resource for deciphering the regulatory architecture underlying immune aging.

immunology↗

bioRxiv2025

IRCAS: a novel end-to-end approach to identify, rectify and classify comprehensive alternative splicing events in a transcriptome without genome reference

Alternative splicing (AS) is a fundamental post-transcriptional mechanism that amplifies proteomic diversity and enables adaptive responses across eukaryotes. Current AS detection methods rely heavily on reference genomes, limiting their applicability to non-model organisms. Existing reference-free approaches suffer from inaccurate splice site prediction and treat detection and classification as separate processes, resulting in cascading errors. We present IRCAS, an integrated end-to-end framework for reference-free AS analysis, comprising three modules: identification, rectification, and classification. IRCAS employs colored de Bruijn graphs for AS detection, an attention-based CNN for splice site rectification, and a hybrid Graph Neural Network combining GAT and Transformer layers for classification. Evaluation across four species demonstrates substantial improvements: splice site accuracy increased to 92-96% versus 50-55% for existing methods, and end-to-end accuracy reached 83.4% compared to 41.2% for the previous best method. IRCAS establishes a new benchmark for reference-free AS detection in non-model organisms. GRAPHICAL ABSTRACT O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=167 SRC="FIGDIR/small/689457v1_fig1.gif" ALT="Figure 1"> View larger version (42K): org.highwire.dtl.DTLVardef@1162531org.highwire.dtl.DTLVardef@e3c2e2org.highwire.dtl.DTLVardef@10f9032org.highwire.dtl.DTLVardef@c3cb5e_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig 1.C_FLOATNO Workflow for construction and application of IRCAS. IRCAS is composed of three parts: identification, rectification, classification. (A) Workflow for reference-free AS identification from a raw transcriptomic data. First, according to the input transcripts, we apply BLAST all versus all alignment for preliminary screen. Then we adopt the MkcDBGAS Graph construction strategy. A cDBG was constructed from two sequences using a specified k-mer size. Based on bubble topologies, bubbles were classified into 5 types: SNV-induced, four AS-induced, MX-induced, AL-induced, AF-induced. (B)Workflow for AS position offset rectification and reconstruction of cDBG. Input transcript pairs are converted into a single sequence that includes two virtual nucleotides denoting the splicing start and end sites. SUPPA, a reference-based method, is utilized to determine the true splicing sites. The sequence is encoded into an nx6 vector using one-hot encoding. The offset between the true and predicted splicing sites is calculated and encoded as the ground truth. An attention-based convolutional neural network (CNN) rectification model is trained on these data to predict the offset, enabling the reconstruction of the cDBG with corrected splicing positions. (C) Workflow for 4 types AS classification. For each cDBG, node features, edge features, and global features are extracted. These features are integrated into distinct layers of a graph attention network (GAT)-Transformer hybrid model. This architecture enables high-precision classification of four types of AS events. (D)Workflow for end-to-end application of IRCAS. Transcriptomic data from any species lacking a reference genome is processed by IRCAS, enabling the classification of seven AS types with high accuracy. C_FIG

bioinformatics↗

bioRxiv2025

From deserts to forests: sequencing of the giant forest hog completes the genomic history of the African Suids

The giant forest hog (GFH, Hylochoerus meinertzhageni), the largest and most elusive wild pig, is found in a disjunct range across equatorial Africa, primarily inhabiting forests and woodlands. It is the only remaining African suid species without whole-genome data. The GFH is currently classified into three subspecies, differing in skull morphology and size. Here, we present the first whole-genome data of nine GFHs; one from Guinea and eight Uganda, representing the two geographically most distant subspecies. This genomic evidence allowed us to resolve the previously uncertain placement of the GFH among African suids as a sister group to the warthogs (Phacochoerus spp.). Also, we found a deep evolutionary split between the two GFH populations resulting in a level of genetic differentiation (FST > 0.75), which is similar to or higher than values observed among species of African wild pigs. The GFH likely diverged from warthogs around 4.6 Mya while the two study populations of GFH diverged at least 0.5 Mya and, subsequently, experienced very dissimilar demographic trajectories. The individual from Guinea had half of the heterozygosity level of the Uganda population, and lower value compared to all the African suids, probably as a result of the long term low Ne and isolation. Due to the close affinity of GFH with the African rainforest biotic zone, our findings have implications for understanding the history of habitat change in Africa. They also indicate a very deep and previously underappreciated evolutionary divergence within the GFH that may have implications for its conservation.

genomics↗

bioRxiv2025

Scalable production of immune-silent circular RNAs for efficient protein expression

Messenger RNA-based therapeutics enable transient protein expression but are limited by its instability. Circular RNA (circRNA) offers enhanced stability, yet scalable production of high-purity circRNA remains challenging. Here, we report an oligo (dT) matrix-based purification strategy for tunable, scalable circRNA production at room temperature. By adjusting poly (A) length, circRNAs are positioned at the midpoint of the oligo (dT) elution profile, enabling one-step separation from both weakly and strongly bound RNA contaminants. We demonstrate that this approach scales linearly with comparable yields for small- and large-scale purifications. In scale-up experiments, 331 mg of highly pure circRNA was recovered in a single run. Incorporating alkaline phosphatase and RNase R digestion prior to chromatography efficiently removes immunogenic contaminants, yielding circRNAs with undetectable immune activation and robust translation in human cells. This workflow provides a practical platform for scalable production of high-quality circRNAs.

bioengineering↗

bioRxiv2025

Biosynthesis of peptidic thiooxazole metallophores installed by multinuclear nonheme iron enzymes.

Significant effort has been directed toward characterization of nonheme iron enzymes owing to their breadth of unique reactivity. Through genome mining, we identified a conserved biosynthetic gene cluster within Pseudomonadota encoding one such family, the multinuclear nonheme iron-dependent oxidative enzymes (MNIO, formerly DUF692). Using a representative gene cluster from Fontimonas thermophila, we heterologously produced the post-translationally modified peptide fontiphorin, and detailed spectral analysis revealed MNIO-catalyzed installation of seven 5-thiooxazole (5TO) moieties. During our work, additional MNIO products were reported with conflicting structural assignments, so we investigated the related biosynthetic gene clusters from Haemophilus influenzae and Neisseria gonorrhoeae. Using alkylation-assisted HMBC correlations, we demonstrated that these products also contain 5TO resulting in a revision of the structure of oxazolin. We further provide evidence supporting a role for 5TO-containing peptides in copper detoxification and recommended this emerging class of Cu-associated peptidic thiooxazole metallophores be referred to as captophorins. To further explore the captophorins, we reconstituted fontiphorin biosynthesis in vitro and investigated its enzymatic requirements. Using cell-free production of single-site, double-site, and naturally occurring variants, we examined enzyme-substrate interactions to determine key sites governing catalysis by 5TO-forming MNIOs. Through our detailed spectroscopic approach for 5TO assignment and investigation of enzyme-substrate interactions, our work unifies tens of thousands of MNIOs in the biosynthesis of captophorins.

biochemistry↗

bioRxiv2025

Multimodal AI Decodes Extreme Environment Functional Dark Matter Beyond Homology

Functional annotation of proteins from extreme environments represents a major bottleneck for bioresource discovery, as a vast reservoir of functional dark matter defies existing homology-based methods. We demonstrate that environmental pressures impart conserved physicochemical energy signatures that co-determine protein function with sequence and structure. Here we developed ACCESS, a multimodal graph neural network employing hierarchical contrastive learning with a tailored label-sample co-embedding to fuse energy, sequence, and structural information and overcome homology scarcity. ACCESS surpasses state-of-the-art methods including BLASTp and CLEAN in annotating low-identity enzymes. Applied to extreme environmental metagenomics, we constructed a function map of extremophile enzymes to expand the biocatalyst library, pinpointed functionally critical residues to guide rational design, and enabled large-scale, function-based macro-evolutionary analyses. This paradigm transcends the limitations of homology, illuminating protein dark matter and accelerating the exploration of the biospheres functional diversity for applications in biotechnology and therapeutic development.

bioinformatics↗

bioRxiv2025

Antibody feedback establishes an affinity brake in the germinal center

Recent advances in mRNA vaccine technology have opened the door to novel types of antigen display, but little is yet known as to how B cells recognize and respond to these formats. By delivering an mRNA-LNP encoded membrane-bound immunogen displaying three conserved HIV-1 Envelope (Env) epitopes to knock-in mouse models with B cell receptors (BCRs) of defined affinities, we investigated how epitope-specific competition shapes germinal center (GC) responses. Co-activation of B cells targeting different epitopes did not alter GC kinetics observed in individual activations, but a striking inverse correlation was observed between BCR affinity and GC residence time in either scenario: high-affinity B cells exhibited shorter persistence in GCs, while those with lower affinity to the antigen were maintained. Furthermore, B cells were able to engage in GC reactions at equivalent rates in the presence or absence of clonal lineages binding the same epitope with similar affinities, while higher-affinity clones suppressed lower-affinity counterparts targeting the same epitope. Spatial transcriptomics revealed plasma-like cells within and adjacent to the GC which, together with the detection of early IgG in draining lymph nodes, suggests that local antibody production from these cells may contribute to feedback-driven kinetics. These findings indicate that a self-modulated local antibody feedback loop may act as a "brake" on epitope-specific recognition--dampening further affinity enhancement for high-affinity B cells and facilitating epitope spreading by redirecting the response toward alternative epitopes.

immunology↗

bioRxiv2025

EMT activates ER-to-Golgi trafficking through upregulation of REEP2 to promote lung cancer progression

Membrane trafficking governs the transport of molecules to both intracellular and extracellular locations, thereby maintaining cell homeostasis. During cancer progression, alterations in membrane trafficking are frequently observed. However, the mechanisms underlying the dysregulation of membrane trafficking in cancer progression remain largely unresolved. Recent evidence has demonstrated that epithelial-to-mesenchymal transition (EMT) in lung adenocarcinoma (LUAD) employs a membrane trafficking program to coordinate cancer cell invasion and immunosuppression in the tumor microenvironment (TME). To further dissect the pro-tumorigenic membrane trafficking program, here we conducted a CRISPR interference (CRISPRi) in vivo screen for membrane trafficking regulators in a syngeneic mouse model with a complete immune system. This screen identified REEP2, an endoplasmic reticulum (ER) shaping protein, as a novel regulator of the EMT-dependent membrane trafficking program, which is associated with a poor prognosis in LUAD patients and is required for LUAD metastasis in a syngeneic orthotopic LUAD mouse model. Mechanistically, the EMT activator ZEB1 upregulates REEP2 expression through miR-183- and miR-193a-mediated regulation that promotes the transportation of secretory cargoes from the ER exit site (ERES) to the Golgi, thereby augmenting the secretion of pro-tumorigenic factors. The REEP2-driven secretion promotes cancer cell proliferation, migration, and the infiltration of myeloid-derived suppressor cells (MDSCs) in the TME. These findings identify REEP2 as a critical mediator of the EMT-driven pro-metastatic membrane trafficking program, revealing a specific vulnerability in mesenchymal LUAD.

cancer biology↗

bioRxiv2025

Simple Methods to Acutely Measure Multiple Timing Metrics among Sexual Repertoire of Male Drosophila

Male Drosophila courtship behavior is a key model for studying temporal decision-making. While courtship index (CI) is widely used to quantify mating activity, other timing metrics like courtship latency, copulation latency (CL), and mating duration (MD) remain understudied. Traditional methods for quantifying these behaviors are often labor-intensive and prone to human error. In this study, we present a protocol combining a modular chamber system and automated software (DrosoMating) to quantify 6 key timing metrics during male courtship. Our image-based video analysis enables precise identification of courtship and copulation events, as well as quantification of their timing and duration under controlled conditions. Validation shows <0.05% error rates and 98-99% agreement with manual scoring for CL, CI, and MD. The protocol supports genetic and neural circuit manipulations, detecting subtle genetic, social, and environmental behavioral variations. By minimizing manual effort and standardizing data collection, this approach facilitates scalable, reproducible studies on adaptive trade-offs, learning, and neural mechanisms in mating behavior. This method streamlines timing analysis in male courtship, offering reproducible metrics for behavioral genetics. HighlightA high-throughput software pipeline for automated temporal profiling of Drosophila melanogaster mating behavior after brief user-guided calibration. An accompanying, open-hardware platform that can be assembled at minimal cost while maintaining experimental rigor. The system attains near-manual accuracy and outputs Temporal Measurement Parameters data that are readily adaptable to--and quantifiable within--diverse behavioral paradigms.

developmental biology↗

bioRxiv2025

A Study on the Pharmacokinetic/Pharmacodynamic Profiles of the Novel PPAR Pan Agonist Chiglitazar Sodium in Rats with Hypoalbuminemia

ObjectiveTo compare the changes in the pharmacokinetic/pharmacodynamic (PK/PD) profiles of Chiglitazar Sodium between healthy rats and those with hypoalbuminemia, and to evaluate the medication safety of Chiglitazar in rats with hypoalbuminemia. MethodsLiquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS) method was employed to determine the plasma concentration ofChiglitazar Sodium in both healthy rats and rats with hypoalbuminemia after administration. Pharmacokinetics were calculated based on the measured data. Blood glucose levels were monitored and recorded in both groups to observe pharmacodynamic changes. ResultsPharmacokinetic analysis revealed that, compared to the healthy group, only the Tmax in the hypoalbuminemia group was significantly earlier by 34.3% (P<0.001), while other pharmacokinetic parameters showed no significant differences. Pharmacodynamic studies indicated that blood glucose levels remained stable in both healthy rats and rats with hypoalbuminemia during the 0.25-10 hour monitoring period. ConclusionExcept for Tmax, there were no significant differences in PK/PD parameters between rats with hypoalbuminemia and healthy rats. Blood glucose levels remained stable in both groups after oral administration of Chiglitazar Sodium. The animal experimental results suggest that, under normal conditions, hypoalbuminemia does not affect the pharmacokinetic and pharmacodynamic profiles of Chiglitazar Sodium.

pharmacology and toxicology↗

bioRxiv2025

Overcoming Extrapolation Challenges of Deep Learning by Incorporating Physics in Protein Sequence-Function Modeling

Understanding protein sequence-to-function relationship is crucial to assist studies of genetic diseases, protein evolution, and protein engineering. The sequence-to-function relationship of proteins is inherently complex due to multi-site high-dimensional correlation and structural dynamics. Deep learning algorithms such as (graph) convolutional neural networks and recently transformers have become very popular for learning the protein sequence-to-function mapping from deep mutational scanning data and available structures. However, it remains very challenging for these models to achieve accurate extrapolation when predicting functional effect of variants with positions or mutation types not seen in the training data. We propose that incorporating the physics of protein interactions and dynamics can be an effective approach to overcome the extrapolation limitations. Specifically, we demonstrate that physics-based modeling can be used to quantify the energetic effects of mutations and that incorporating these physical energetics directly within the convolution and graph convolution neural networks can significantly improve the performance of positional and mutational extrapolation compared to models without biophysics-inspired features. Our results support the effectiveness of leveraging physical knowledge in overcoming the limitation of data scarcity.

bioinformatics↗

bioRxiv2025

Missense Mutation ADNP p.C687R Disrupts Chromatin Regulation and GABAergic Differentiation in HVDAS

Activity-dependent neuroprotective protein (ADNP) is a key regulator of neurodevelopment and a high-risk gene for autism spectrum disorder (ASD). Most pathogenic variants cause loss-of-function in Helsmoortel-Van der Aa syndrome (HVDAS), yet the impact of missense mutations remains unclear. Here, we characterized a ADNP missense mutation (p.C687R) identified from a HVDAS patient, predicted to disrupt its ninth zinc finger domain. Overexpression of p.C687R altered subnuclear localization and affected wild-type ADNP distribution. In utero electroporation revealed impaired neuronal migration and cortical branching. CUT&Tag profiling showed mutant- specific chromatin binding, preferentially targeting histone modification genes through both direct and distal interactions. Patient-derived iPSCs carrying p.C687R exhibited upregulated bivalent histone marks (H3K4me3 and H3K27me3) on neurodevelopmental genes, including regulators of GABAergic differentiation, which became transcriptionally activated in neural progenitors. These findings reveal a gain-of-function mechanism by which ADNP p.C687R perturbs chromatin regulation and disrupts GABAergic lineage specification, expanding the molecular framework of ADNP pathogenesis. The Paper Explained ProblemMutations in the gene ADNP cause Helsmoortel-Van der Aa syndrome (HVDAS), a severe neurodevelopmental disorder, and also increase the risk for autism spectrum disorder (ASD). While most known ADNP mutations lead to loss of protein function, many missense mutations that alter a single amino acid remain poorly understood regarding their effects on brain development and disease contribution. ResultsWe characterized a specific missense mutation, ADNP p.C687R, from a patient diagnosed with HVDAS, and found that it does not simply inactivate the protein but instead confers a disruptive new function. Across multiple models, from cells to mice, the p.C687R mutant exhibits abnormal chromatin binding, altered subnuclear localization, and impaired neuronal migration and differentiation. Importantly, in patient-derived iPSCs, it marks distinct neurodevelopmental genes with bivalent histone modifications, which become aberrantly activated during differentiation into GABAergic neural progenitor cells. ImpactThese findings confirm the pathogenicity of the missense mutation ADNP p.C687R in HVDAS, reveal a gain-of-function mechanism whereby the mutant disrupts chromatin regulation and GABAergic lineage specification, and advance our understanding of ADNP-related neurodevelopmental disorders.

neuroscience↗

bioRxiv2025

Mechanistic insights into Phage SSB-activated bacterial Retron-Eco8 immunity

The Retron-Eco8 system, comprising a reverse transcriptase (RT), a non-coding RNA (ncRNA), and an OLD-family nuclease effector, protects bacteria from phage infection via abortive infection upon sensing a phage single-stranded DNA-binding protein (SSB). However, the molecular basis of this immunity remained unclear. Here, we report cryo-electron microscopy (cryo-EM) structures of Retron-Eco8 in inactive and activated states, revealing mechanisms of phage-triggered activation and effector function. Retron-Eco8 assembles into a tetrameric complex in which each protomer contains an RT, msrRNA-msdDNA duplex, and effector in an autoinhibited conformation. Upon phage infection, phage SSB binds msdDNA, relieving autoinhibition and activating the nuclease effector to degrade both phage and host DNA, triggering cell death to block phage propagation. Host SSB fails to activate the system, while DNA binding and oligomerization of phage SSB are essential for this activation, highlighting its specificity. These findings elucidate the molecular mechanism of Retron-Eco8-mediated immunity, facilitating retron-based biotechnological applications.

microbiology↗