Biology subjects

Dhakal, A.

Publications and source records attributed to Dhakal, A..

12 recordsLinked to original sources

bioRxiv2026

Does Short Molecular Dynamics Rescoring Improve the Ranking of p38α Inhibitors over Docking? A Replicate Benchmark

Rescoring docked ligands with molecular dynamics (MD) and end point free energy methods such as MM/GBSA is widely used, but its benefit over docking for series of close analogs is uncertain. We compared AutoDock Vina scores with Md based MM/GBSA energies for 29 neutral p38 kinase inhibitors (PDB 3FLY) from the Open Force Field protein-ligand benchmark, whose IC50 values from a single publication span 3.8 kcal/mol in estimate {Delta}G. Each ligand was docked with three random seeds and simulated in two independent 20 ns replicas started from the benchmark pose (AMBER99SB-ILDN, GAFF2, TIP3L; GROMACS). MM/GBSA energies were calculated from 251 frames per trajectory without an entropy term. All 58 simulations remained stable with the ligand bound. Vina scores correlated weakly with experimental {Delta}G (Spearman {rho} = 0.28 98% CI -0.15 to 0.62), as did replica averaged MM/GBSA energies ({rho} = 0.24 -0.18 to 0.62). The difference was -0.03 (-0.50 to 0.43), so the methods were not distinguishable. Independent replicas of the same ligand differed by 3.3 kcal/mol on average, comparable to the experimental range. Raising the interior dielectric constant from 1 to 2 gave a nearly identical ranking ({rho} = 0.96 between the two sets of scores; {rho} = 0.37 against experiment, -0.04 to 0.69). In an exploratory analysis, ligands size alone correlated with affinity ({rho} = 0.60, 0.28 to 0.82), whereas neither computed score tracked size. For this narrow range series, short MD rescoring gave no measurable improvement over docking, and replica variability limited its resolution.

biophysics↗

bioRxiv2026

TomoSwin3D: a Swin3D Transformer for the Identification and Classification of Macromolecules in 3D Cryo-ET Tomograms

Cryo-electron tomography (cryo-ET) enables in situ three-dimensional visualization of many protein complexes and other macromolecular assemblies such as ribosomes in cells, yet automated macromolecule particle identification in 3D cryo-ET tomograms remains a major bottleneck due to dose-limited low signal-to-noise ratios, missing-wedge artifacts, and densely crowded cellular backgrounds. We present TomoSwin3D, an end-to-end three-dimensional (3D) macromolecule particle identification and classification pipeline centered on a Swin Transformer-based U-Net that performs particle identification and classification and outputs particle centroid coordinates. TomoSwin3D leverages a multi-channel input representation that augments raw tomogram densities with complementary 3D feature maps capturing edge strength (Sobel gradients), local contrast enhancement (morphological top-hat), and multiscale blob responses (Difference-of-Gaussians), improving detectability of small and low-contrast targets. To better preserve particle geometry and avoid hand-crafted shape assumptions, it adopts occupancy-preserving supervision that directly uses available 3D instance masks rather than heuristic Gaussian/spherical labels and applies scalable patch-wise inference followed by lightweight post-processing (connected-component analysis, size filtering, centroid extraction) for robust centroid coordinate extraction. Across diverse simulated and experimental cryo-ET tomogram benchmarks including SHREC 2021 and 2020 test datasets, EMPIAR dataset, and Cryo-ET data portal dataset, TomoSwin3D achieves strong and consistent performance in detecting proteins and other particles, outperforming existing methods, with a pronounced advantage in picking hard, small protein particles. These results establish TomoSwin3D as a scalable and accurate solution for high-throughput cryo-ET macromolecule particle picking and downstream subtomogram averaging.

biochemistry↗

bioRxiv2025

CryoFSL: an annotation-efficient, few-shot learning framework for robust protein particle picking in cryo-EM micrographs

AO_SCPLOWBSTRACTC_SCPLOWAccurate identification of protein particles in cryo-electron microscopy (cryo-EM) micrographs is crucial for high-resolution structure determination, but remains challenging due to the heavy reliance on extensive annotated datasets and the difficulty of ensuring robustness under low signal-to-noise ratio (SNR) conditions. Current approaches require large annotations and exhibit poor generalization to new protein targets. We present CryoFSL (Cryo-EM Few Shot-Learning), a novel few-shot learning framework built upon Segment Anything Model 2 (SAM2) with lightweight adapters, enabling robust particle picking using as few as five labeled micrographs, significantly reducing annotation burden. The frameworks hierarchical adapter design supports dynamic feature modulation for low-SNR and heterogeneous conditions, resolving the trade-off between annotation burden and performance. CryoFSL surpasses both traditional template-based methods and state-of-the-art deep learning models across diverse proteins in the few-short learning setting, achieving superior recall, precision and 3D reconstruction resolution with minimal supervision. It maintains stability across heterogeneous micrographs and consistently detects high-quality particles with fewer false positives. Notably, CryoFSL achieves competitive density map reconstruction resolution with just a fraction of the particles picked by other methods, redefining efficiency and quality in cryo-EM analysis. This work paves the way for scalable, generalizable, and annotation-efficient particle picking pipelines. The code is available at GitHub.

bioinformatics↗

bioRxiv2025

Multimodal deep learning integration of cryo-EM and AlphaFold3 for high-accuracy protein structure determination

Cryo-electron microscopy (cryo-EM) is a key technology for determining the structures of proteins, particularly large protein complexes. However, automatically building high-accuracy protein structures from cryo-EM density maps remains a crucial challenge. In this work, we introduce MICA, a fully automatic and multimodal deep learning approach combining cryo-EM density maps with AlphaFold3-predicted structures at both input and output levels to improve cryo-EM protein structure modeling. It first uses a multi-task encoder-decoder architecture with a feature pyramid network to predict backbone atoms, C atoms and amino acid types from both cryo-EM maps and AlphaFold3-predicted structures, which are used to build an initial backbone model. This model is further refined using AlphaFold3-predicted structures and density maps to build final atomic structures. MICA significantly outperforms other state-of-the-art deep learning methods in terms of both modeling accuracy and completeness and is robust to protein size and map resolution. Additionally, it builds high-accuracy structural models with an average template-based modeling score (TM-score) of 0.93 from recently released high-resolution cryo-EM density maps, showing it can be used for real-world, automated, accurate protein structure determination.

bioinformatics↗

bioRxiv2023

CryoVirusDB: A Labeled Cryo-EM Image Dataset for AI-Driven Virus Particle Picking

With the advancements in instrumentation, image processing algorithms, and computational capabilities, single-particle electron cryo-microscopy (cryo-EM) has achieved nearly atomic resolution in determining the 3D structures of viruses. The virus structures play a crucial role in studying their biological function and advancing the development of antiviral vaccines and treatments. Despite the effectiveness of artificial intelligence (AI) in general image processing, its development for identifying and extracting virus particles from cryo-EM micrographs (images) has been hindered by the lack of manually labelled high-quality datasets. To fill the gap, we introduce CryoVirusDB, a labeled dataset containing the coordinates of expert-picked virus particles in cryo-EM micrographs. CryoVirusDB comprises 9,941 micrographs of 9 different viruses along with the coordinates of 339,398 labeled virus particles. It can be used to train and test AI and machine learning (e.g., deep learning) methods to accurately identify virus particles in cryo-EM micrographs for building atomic 3D structural models for viruses.

bioinformatics↗

bioRxiv2023

CryoTransformer: A Transformer Model for Picking Protein Particles from Cryo-EM Micrographs

Cryo-electron microscopy (cryo-EM) is a powerful technique for determining the structures of large protein complexes. Picking single protein particles from cryo-EM micrographs (images) is a crucial step in reconstructing protein structures from them. However, the widely used template-based particle picking process requires some manual particle picking and is labor-intensive and time-consuming. Though machine learning and artificial intelligence (AI) can potentially automate particle picking, the current AI methods pick particles with low precision or low recall. The erroneously picked particles can severely reduce the quality of reconstructed protein structures, especially for the micrographs with low signal-to-noise (SNR) ratios. To address these shortcomings, we devised CryoTransformer based on transformers, residual networks, and image processing techniques to accurately pick protein particles from cryo-EM micrographs. CryoTransformer was trained and tested on the largest labelled cryo-EM protein particle dataset - CryoPPP. It outperforms the current state-of-the-art machine learning methods of particle picking in terms of the resolution of 3D density maps reconstructed from the picked particles as well as F1-score and is poised to facilitate the automation of the cryo-EM protein particle picking.

bioinformatics↗

bioRxiv2023

Accurate cryo-EM protein particle picking by integrating the foundational AI image segmentation model and specialized U-Net

Picking protein particles in cryo-electron microscopy (cryo-EM) micrographs is a crucial step in the cryo-EM-based structure determination. However, existing methods trained on a limited amount of cryo-EM data still cannot accurately pick protein particles from noisy cryo-EM images. The general foundational artificial intelligence (AI)-based image segmentation model such as Metas Segment Anything Model (SAM) cannot segment protein particles well because their training data do not include cryo-EM images. Here, we present a novel approach (CryoSegNet) of integrating an attention-gated U-shape network (U-Net) specially designed and trained for cryo-EM particle picking and the SAM. The U-Net is first trained on a large cryo-EM image dataset and then used to generate input from original cryo-EM images for SAM to make particle pickings. CryoSegNet shows both high precision and recall in segmenting protein particles from cryo-EM micrographs, irrespective of protein type, shape, and size. On several independent datasets of various protein types, CryoSegNet outperforms two top machine learning particle pickers crYOLO and Topaz as well as SAM itself. The average resolution of density maps reconstructed from the particles picked by CryoSegNet is 3.32 [A], 7% better than 3.57 [A] of Topaz and 14% better than 3.85 [A] of crYOLO.

bioinformatics↗

bioRxiv2023

Predicting Protein-Ligand Binding Structure Using E(n) Equivariant Graph Neural Networks

Drug design is a costly and time-consuming process, often taking more than 12 years and costing up to billions of dollars. The COVID-19 pandemic has signified the urgent need for accelerated drug development. The initial stage of drug design involves the identification of ligands that exhibit a strong affinity for specific binding sites on protein targets (receptors), along with the determination of their precise binding conformation (3-dimensional (3D) structure). However, accurately determining the 3D conformation of a ligand binding with its target remains challenging due to the limited capability of exploring the huge chemical and protein structure space. To address this challenge, we propose a new E(n) Equivariant Graph Neural Network (EGNN) method for predicting the 3D binding structures of ligands and proteins. By treating proteins and ligands as graphs, the method extracts residue/atom-level node and edge features and utilizes physicochemical and geometrical properties of proteins and ligands to predict their binding structures. The results demonstrate the promising potential of EGNN for predicting ligand-protein binding poses.

biochemistry↗

bioRxiv2023

ADARs employ a neural-specific mechanism to regulate PQM-1 expression and survival from hypoxia

The ability to alter gene expression programs in response to changes in environmental conditions is central to the ability of an organism to thrive. For most organisms, the nervous system serves as the master regulator in communicating information about the animals surroundings to other tissues. The information relay centers on signaling pathways that cue transcription factors in a given cell type to execute a specific gene expression program, but also provide a means to signal between tissues. The transcription factor PQM-1 is an important mediator of the insulin signaling pathway contributing to longevity and the stress response as well as impacting survival from hypoxia. Herein, we reveal a novel mechanism for regulating PQM-1 expression specifically in neural cells of larval animals. Our studies reveal that the RNA binding protein, ADR-1, binds to pqm-1 mRNA in neural cells. This binding is regulated by the presence of a second RNA binding protein, ADR-2, which when absent leads to reduced expression of both pqm-1 and downstream PQM-1 activated genes. Interestingly, we find that neural pqm-1 expression is sufficient to impact gene expression throughout the animal and affect survival from hypoxia; phenotypes that we also observe in adr mutant animals. Together, these studies reveal an important post-transcriptional gene regulatory mechanism that allows the nervous system to sense and respond to environmental conditions to promote organismal survival from hypoxia.

genetics↗

bioRxiv2023

ADARs regulate cuticle collagen expression to promote survival to pathogen infection

BackgroundIn all organisms, the innate immune system defends against pathogens through basal expression of molecules that provide critical barriers to invasion and inducible expression of effectors that combat infection. The adenosine deaminase that act on RNA (ADAR) family of RNA binding proteins has been reported to influence innate immunity in metazoans. However, studies on the susceptibility of ADAR mutant animals to infection are largely lacking. ResultsHere, by analyzing adr-1 and adr-2 null mutants in well-established slow-killing assays, we find that both Caenorhabditis elegans ADARs are important for organismal survival to gram-negative and gram-positive bacteria, all of which are pathogenic to humans. Furthermore, our high-throughput sequencing and genetic analysis reveal that ADR-1 and ADR-2 function in the same pathway to regulate collagen expression. Consistent with this finding, our scanning electron microscopy studies indicate adr-1;adr-2 mutant animals also have altered cuticle morphology prior to pathogen exposure. ConclusionsOur data uncover a critical role of the C. elegans ADAR family of RNA binding proteins in promoting cuticular collagen expression, which represents a new post-transcriptional regulatory node that influences the extracellular matrix. In addition, we provide the first evidence that ADAR mutant animals have altered susceptibility to infection with several opportunistic human pathogens, suggesting a broader role of ADARs in altering physical barriers to infection to influence innate immunity.

molecular biology↗

bioRxiv2023

CryoPPP: A Large Expert-Labelled Cryo-EM Image Dataset for Machine Learning Protein Particle Picking

Cryo-electron microscopy (cryo-EM) is currently the most powerful technique for determining the structures of large protein complexes and assemblies. Picking single-protein particles from cryo-EM micrographs (images) is a key step in reconstructing protein structures. However, the widely used template-based particle picking process is labor-intensive and time-consuming. Though the emerging machine learning-based particle picking can potentially automate the process, its development is severely hindered by lack of large, high-quality, manually labelled training data. Here, we present CryoPPP, a large, diverse, expert-curated cryo-EM image dataset for single protein particle picking and analysis to address this bottleneck. It consists of manually labelled cryo-EM micrographs of 32 non-redundant, representative protein datasets selected from the Electron Microscopy Public Image Archive (EMPIAR). It includes 9,089 diverse, high-resolution micrographs ([~]300 cryo-EM images per EMPIAR dataset) in which the coordinates of protein particles were labelled by human experts. The protein particle labelling process was rigorously validated by both 2D particle class validation and 3D density map validation with the gold standard. The dataset is expected to greatly facilitate the development of machine learning and artificial intelligence methods for automated cryo-EM protein particle picking. The dataset and data processing scripts are available at https://github.com/BioinfoMachineLearning/cryoppp

bioinformatics↗

bioRxiv2021

Functional differences drive the selection of NRAS mutants in melanoma

Distinct NRAS mutants are enriched in various tumor types. Here, we generated a suite of fully congenic, conditional, Nras knock-in mouse models (LSL-Nras Q61R, -K, -L, -H, -P, -Q; G12D and G13D, -R) to test the hypothesis that melanocyte transformation requires functions specific to the NRAS mutants enriched in human melanoma (Q61R and Q61K). Consistent with the rarity of NRAS codon 12 and 13 mutants in human melanoma, spontaneous melanomas were rare or absent in mice expressing NRAS G12D, G13D or G13R. Mice expressing less common codon 61 alleles (Q61H, Q61P) also developed few or no tumors. NRAS Q61R, Q61K, or Q61L expression, by contrast, induced rapid melanoma onset with high penetrance. Cohorts of heterozygous mice containing one LSL-Nras Q61R and one LSL-Nras Q61K, -L, -H, -P, or -Q allele were generated to assess potential interactions between NRAS mutants. The ability of each Nras variant to substitute for an Nras Q61R allele was consistent with its own ability to drive spontaneous melanoma formation. However, LSL-Nras Q61Q/Q61R mice rarely developed tumors. In vitro experiments in mouse embryonic fibroblasts (MEFs) highlighted activation of the MAPK pathway as a defining difference between tumorigenic and non-tumorigenic NRAS mutants. Enhanced MAPK activation was associated with the promotion of BRAF-BRAF and BRAF-CRAF dimers. These results support the development of cancer preventative strategies specific to the properties of the commonly observed RAS mutants in each tumor type.

cancer biology↗