Biology subjects

Teng, D.

Publications and source records attributed to Teng, D..

12 recordsLinked to original sources

bioRxiv2026

MG2Act: A Mechanism-Inspired Sequential Attention Framework for Molecular Glue Degradation Prediction

Molecular glue degraders act by inducing productive proximity between an E3 ligase and a substrate protein. For most characterized degradative glues, a small molecule first engages the E3, conditions its substrate-recognition surface, and only then enables recruitment of a compatible neo-substrate. This directionality is rarely encoded explicitly in computational models, which typically fuse molecule, E3 and target representations simultaneously. We present MG2Act, a structure-independent framework that translates this two-step logic into sequential cross-attention, using CRBN-mediated degradation as the most data-rich representative system. Starting from a curated continuous-valued benchmark of 1,207 pairs across 47 targets, a refined subset of 1,159 pairs was selected to train MG2Act after excluding rare targets. On identical processed data, MG2Act consistently outperforms machine learning baselines, robustly generalizes under strict redundancy-filtering, and responds coherently to mechanism-based perturbations. Prospective screening and zero-shot target-conditioned prioritization identified nanomolar degraders of IKZF1, CK1 and CDK4, including the non-classical IMiD-core CDK4 degrader SWC-202.

bioinformatics↗

bioRxiv2026

Functional orthogonality of WRN inhibitor resistance enables alternating therapy and mutation-tolerant inhibitor design in MSI-H cancers

Werner syndrome helicase (WRN) is a synthetic-lethal vulnerability in microsatellite instability-high (MSI-H)/mismatch repair-deficient (dMMR) cancers1-5, and non-covalent and covalent WRN inhibitors are now entering clinical development6-9. A central unresolved question is whether resistance to one WRN inhibitor class inevitably compromises the target, or instead creates actionable vulnerabilities to an alternate modality. Here we generated ten stable acquired-resistance models across three MSI-H cell lines using the non-covalent inhibitor HRO761 and the covalent inhibitor VVD-214. Whole-exome sequencing, biochemical reconstitution and isogenic knock-in models identified recurrent on-target WRN missense mutations as dominant resistance drivers, but with sharply modality-specific spectra: HRO761 resistance clustered at G729/F730/I852, whereas VVD-214 resistance concentrated at E846. These mutations impaired inhibitor engagement and abolished the canonical WRN inhibitor (WRNi)-induced pharmacodynamic cascade, including WRN reduction, DNA damage response (DDR) activation and G2/M arrest. Crucially, most resistance mutations retained biochemical, cellular and in vivo sensitivity to the alternate inhibitor modality, revealing a functional orthogonality that enabled a 7-day cyclic alternating regimen to delay tumour regrowth in xenografts. We further identified F730L as an engineered cross-resistant bottleneck model and used structure-guided, artificial intelligence (AI)-enabled optimization to generate GBA-007, a proof-of-concept mutation-tolerant WRN inhibitor candidate with promising activity against F730L in vitro and in vivo. Thus, clinically relevant WRN inhibitor classes impose distinct on-target resistance trajectories that can be exploited through schedule design, while cross-resistant bottlenecks can be addressed by rapid mutation-aware inhibitor engineering.

Cancer Biology↗

bioRxiv2026

MAHLER: Integrating Metadynamics and Inverse Folding to Predict Antibody-Antigen Kinetics

Binding kinetics are crucial for antibody function, shaping pharmacokinetics and in vivo efficacy beyond what equilibrium affinity captures. We present "Metadynamics-Anchored Hybrid Learning for Engineering off-Rates (MAHLER)", a fully open-source machine learning/physics hybrid method that predicts relative antibody-antigen residence times at scale. Incorporating inverse-folding models into molecular dynamics simulations, MAHLER shows first-in-class screening-grade accuracy in calculating relative antibody-antigen dissociation kinetics across a family of point mutants. After initial antigen-specific setup, each prediction takes only 4 minutes on a single NVIDIA A100 GPU, compared to days even with already enhanced molecular dynamics simulations. This provides practical kinetics-aware complement to current computational design approaches that focus primarily on binding affinity for antibody-antigen complexes.

biophysics↗

bioRxiv2026

Detecting and quantifying overparametrization in RNA language models with REDIAL

While RNA language models (LMs) have served as foundation models (FMs) to advanced structural prediction, their evaluation relies heavily on supervised downstream tasks. Such tasks can often mask FM inefficiencies and reflect downstream training set memorization. To address this, here we introduce REDIAL (RNA Embedding perturbation Diagnostics for Language models), a zero-shot, unsupervised framework designed to extract coevolutionary signals directly from the high-dimensional latent spaces of RNA language models. By applying REDIAL, we uncover stark, layer-wise disparities in how popular RNA language models (LMs) internalize structural constraints through a layer-wise dissection and ablation study. Our results showed how such layerwise behavior deviates from protein LMs and is related to design flaws in the architectures. Specifically, we show that current RNA LMs are severely overparameterized relative to the limited sequence diversity of available RNA databases, leading to profound parameter inefficiency and overfitting. Furthermore, we establish that structure-guided pretraining fundamentally improves the signal-to-noise ratio of learned coevolutionary couplings compared to sequence-only baselines. Ultimately, this unsupervised evaluation paradigm exposes critical flaws in current parameter scaling strategies and provides a rigorous diagnostic benchmark to guide the development of more efficient, generalizable foundation models for RNA therapeutics and de novo design.

bioinformatics↗

bioRxiv2025

Benchmarking Real-World Applicability of Molecular Generative Models from De novo Design to Lead Optimization with MolGenBench

Structure-based drug design (SBDD) has been profoundly reshaped by the advent of deep generative models, yet their practical impact on drug discovery remains limited. A central issue is the absence of a rigorous, application-oriented benchmark that mirrors the multi-stage, target-aware workflows of real-world pharmaceutical development. Inspired by recent advances in benchmarking for computer vision and large language models, where systematic evaluation has catalysed rapid progress, we introduce MolGenBench, a comprehensive benchmark designed to close the gap between molecular generation algorithms and tangible drug discovery outcomes. MolGenBench integrates a structurally diverse, large-scale dataset spanning 120 protein targets, 5,433 chemical series comprising 220,005 experimentally confirmed active molecules. Beyond conventional de novo generation, it incorporates a dedicated hit-to-lead (H2L) optimization scenario, which represents a critical phase in hit optimization that is seldom evaluated in existing benchmarks. We further introduce novel, pharmaceutically grounded metrics that assess a models ability to both rediscover target-specific actives or progressively optimize compounds for potency. Through extensive evaluation, MolGenBench uncovers significant gaps between current generative models and the demands of real-world drug development, establishing a foundational resource for building generative models with enhanced practical impact and accelerated translational potential.

bioinformatics↗

bioRxiv2025

Advancing Ligand Binding Affinity Prediction with Cartesian Tensor-Based Deep Learning

0.We present PBCNet2.0, a cartesian tensor-based Siamese Neural Network for protein-ligand relative binding affinity prediction. Trained on 8.6 million protein-ligand complex structure pairs, PBCNet2.0 achieves zero-shot performance comparable to computationally intensive physics-based simulations. Our prioritization experiments show that PBCNet2.0 speeds up binding affinity optimization by 718% while reducing resource use by 41%. Through extensive retrospective experiments, we demonstrate that PBCNet2.0 intrinsically comprehends protein-ligand interactions, showing high sensitivity to intermolecular interactions and exceptional perception of spatial geometric information. Strikingly, PBCNet2.0 exhibits an emergent capability to predict affinity changes induced by binding residue variations, highlighting its potential for identifying resistance mutation. We prospectively validated these capabilities on two targets ENPP1 and ALDH1B1, where PBCNet2.0 successfully identified affinity shifts arising from subtle molecular interactions and conformational differences, and pinpointed critical binding residues with an 83% hit rate. This combination of computational efficiency, spatial geometric perception of binding site, and generalizable affinity prediction establishes PBCNet2.0 as a transformative tool for developing pharmacological probes for all human proteins.

molecular biology↗

bioRxiv2024

The C-terminus of the multi-drug efflux pump EmrE prevents proton leak by gating transport

The model multi-drug efflux pump from Escherichia coli, EmrE, can perform multiple types of transport leading to different biological outcomes, conferring resistance to some drug substrates and enhancing susceptibility to others. While transporters have traditionally been classified as antiporters, symporters, or uniporters, there is growing recognition that some transporters may exhibit mixed modalities. This raises new questions about the regulation and mechanisms of these transporters. Here we show that the C-terminal tail of EmrE acts as a secondary gate, preventing proton leak in the absence of drug. Substrate binding unlocks this gate, allowing transport to proceed. Truncation of the C-terminal tail ({Delta}107-EmrE) leads to altered pH regulation of alternating access, an important kinetic step in the transport cycle, as measured by NMR. {Delta}107-EmrE has increased proton leak in proteoliposome assays and bacteria expressing this mutant have reduced growth. MD simulations of {Delta}107-EmrE show formation of a water wire from the open face of the transporter to the primary binding site in the core, facilitating proton leak. In WT-EmrE, the C-terminal tail forms specific interactions that block formation of the water wire. Together these data strongly support the C-terminus of EmrE acting as a secondary gate that regulates access to the primary binding site in the core of the transporter.

biophysics↗

bioRxiv2024

Activation of the influenza B M2 proton channel (BM2)

Influenza B viruses have co-circulated during most seasonal flu epidemics and can cause significant human morbidity and mortality due to their rapid mutation, emerging drug resistance, and severe impact on vulnerable populations. The influenza B M2 proton channel (BM2) plays an essential role in viral replication, but the mechanisms behind its symmetric proton conductance and the involvement of a second histidine (His27) cluster remain unclear. Here we perform the membrane-enabled continuous constant-pH molecular dynamics simulations on wildtype BM2 and a key H27A mutant to explore its pH-dependent conformational switch. Simulations capture the activation as the first histidine (His19) protonates and reveal the transition at lower pH values compared to AM2 is a result of electrostatic repulsions between His19 and pre-protonated His27. Crucially, we provide an atomic-level understanding of the symmetric proton conduction by identifying pre-activating channel hydration in the C-terminal portion. This research advances our understanding of the function of BM2 function and lays the groundwork for further chemically reactive modeling of the explicit proton transport process as well as possible anti-flu drug design efforts.

biophysics↗

bioRxiv2022

Real-Time Spatial Registration for 3D Human Atlas

AO_SCPLOWBSTRACTC_SCPLOWThe human body is made up of about 37 trillion cells (adults). Each cell has its own unique role and is affected by its neighboring cells and environment. The NIH Human BioMolecular Atlas Program (HuBMAP) aims at developing a 3D atlas of human body consisting of organs, vessels, tissues to singe cells with all 3D spatially registered in a single 3D human atlas using tissues obtained from normal individuals across a wide range of ages. A critical step of building the atlas is to register 3D tissue blocks in real-time to the right location of a human organ, which itself consists of complex 3D substructures. The complexity of the 3D organ model, e.g., 35 meshes for a typical kidney, poses a significant computational challenge for the registration. In this paper, we propose a comprehensive framework TICKET (TIssue bloCK rEgisTration) for tissue block registration for 3D human atlas, including (1) 3D mesh pre-processing, (2) spatial queries on intersection relationship and (3) intersection volume computation between organs and tissue blocks. To minimize search space and computation cost, we develop multi-level indexing on both the anatomical structure level and mesh level, and utilize OpenMP for parallel computing. Considering cuboid based shape of the tissue block, we propose an efficient voxelization-based method to estimate the intersection volume. Our experiments demonstrate that the proposed framework is practical and efficient. TICKET is now integrated into the HuBMAP CCF registration portal [1].

bioinformatics↗

bioRxiv2022

Transretinal migration of astrocytes and brain/spinal cord-like cells arising from transplanted human retinal organoids.

Human retinal organoid transplantation can potentially restore vision in patients with degenerative retinal diseases. How the recipient retina regulates the maturation, fate specification, and migration of transplanted organoid cells is unknown. We transplanted human retinal organoid-derived cells into photoreceptor-deficient mice, conducted histology and single-cell RNA sequencing analyses, and observed two main classes of graft-derived cells. The first class consisted of retinal astrocytes and brain/spinal cord-like neural precursors, absent or rare in cultured organoids, that migrated into all recipient retinal layers and traveled long distances. The second class consisted of retinal progenitor-derived cells, including rods and cones, that remained in the subretinal space and matured more rapidly than photoreceptors in culture. These data suggest that the recipient subretinal space promotes the maturation of transplanted photoreceptors while inducing or expanding migratory cell populations that are not normally derived from retinal progenitors. These findings have important implications for cell-based treatment of retinal diseases.

neuroscience↗

bioRxiv2019

Quantifiable In Vivo Imaging Biomarkers of Retinal Regeneration by Photoreceptor Cell Transplantation

PurposeShort-term improvements in retinal anatomy are known to occur in preclinical models of photoreceptor transplantation. However, correlative changes over the long term are poorly understood. We aimed to develop a quantifiable imaging biomarker grading scheme, using non-invasive multimodal confocal scanning laser ophthalmoscopy (cSLO) imaging, to enable serial evaluation of photoreceptor transplantation over the long term. MethodsYellow-green fluorescent microspheres were transplanted into the vitreous cavity and/or subretinal space of C57/BL6J mice. Photoreceptor cell suspensions or sheets from rhodopsin-green fluorescent protein mice were transplanted subretinally, into either NOD.CB17-Prkdcscid/J or C3H/HeJ-Pde6brd1 mice. Multimodal cSLO imaging was performed serially for up to three months after transplantation. Imaging biomarkers were scored, and a grade was defined for each eye by integrating the scores. Image grades were correlated with immunohistochemistry (IHC) data. ResultsMultimodal imaging enabled the extraction of quantitative imaging biomarkers including graft size, GFP intensity, graft length, on-target graft placement, intra-graft lamination, hemorrhage, retinal atrophy, and peri-retinal proliferation. Migration of transplanted material was observed. Changes in biomarker scores and grades were detected in 13/16 and 7/16 eyes, respectively. A high correlation was found between image grades and IHC parameters. ConclusionsSerial evaluation of multiple imaging biomarkers, when integrated into a per-eye grading scheme, enabled comprehensive tracking of longitudinal changes in photoreceptor cell grafts over time. The application of systematic multimodal in vivo imaging could be useful in increasing the efficiency of preclinical retinal cell transplantation studies in rodents and other animal models.

neuroscience↗

bioRxiv2019

Spread of pathological α-synuclein from urogenital nerves initiates multiple system atrophy-like symptoms

Multiple system atrophy (MSA) is a fatal adult-onset movement disorder with autonomic failures, especially urogenital dysfunction. The neuropathological feature of MSA is the accumulation of misfolded -synuclein (-Syn) in the nervous system. Here, we show that misfolded -Syn exist in nerve terminals in detrusor (DET) and external urethral sphincter (EUS) of patients with MSA. Moreover, -Syn preformed fibrils inoculated into the EUS or DET in TgM83+/- mice initiated the transmission of misfolded -Syn from the lower urinary tract to brain, and these mice developed -Syn inclusion pathology through micturition reflex pathways along with urinary dysfunction and motor impairments. These findings indicate that spreading of misfolded -Syn from the autonomic control of the lower urinary tract to the brain via micturition reflex pathways induces autonomic failure and motor impairments. These results provide important new insights into the pathogenesis of MSA as well as highlight potential targets for early detection and therapeutics.

neuroscience↗