Biology subjects

Hy, T. S.

Publications and source records attributed to Hy, T. S..

11 recordsLinked to original sources

bioRxiv2025

LANTERN: Leveraging Large Language Models and Transformers for Enhanced Molecular Interactions

Understanding molecular interactions such as Drug-Target Interaction (DTI), Protein-Protein Interaction (PPI), and Drug-Drug Interaction (DDI) is critical for advancing drug discovery and systems biology. However, existing methods often struggle with scalability due to the vast chemical and biological space and suffer from limited accuracy when capturing intricate biochemical relationships. To address these challenges, we introduce LANTERN (Leveraging Large LANguage Models and Transformers for Enhanced moleculaR interactioNs), a novel deep learning framework that integrates Large Language Models (LLMs) with Transformer-based architectures to model molecular interactions more effectively. LANTERN generates high-quality, context-aware embeddings for drug and protein sequences, enabling richer feature representations and improving predictive accuracy. By leveraging a Transformer-based fusion mechanism, our framework enhances scalability by efficiently integrating diverse interaction data while maintaining computational feasibility. Experimental results demonstrate that LANTERN achieves state-of-the-art performance on multiple DTI and DDI benchmarks, significantly outperforming traditional deep learning approaches. Additionally, LANTERN exhibits competitive performance on challenging PPI tasks, underscoring its versatility across diverse molecular interaction domains. The proposed framework offers a robust and adaptable solution for modeling molecular interactions, efficiently handling a diverse range of molecular entities without the need for 3D structural data and making it a promising framework for foundation models in molecular interaction. Our findings highlight the transformative potential of combining LLM-based embeddings with Transformer architectures, setting a new standard for molecular interaction prediction. The source code and relevant documentation are available at: https://github.com/HySonLab/LANTERN.

bioinformatics↗

bioRxiv2024

Generative AI-assisted Virtual Screening Pipeline for Generalizable and Efficient Drug Repurposing

Drug repurposing presents a valuable strategy to expedite drug discovery by identifying new therapeutic uses for existing compounds, especially for diseases with limited treatment options. We propose a Generative AI-assisted Virtual Screening Pipeline that combines generative modeling, binding pocket prediction, and similarity-based searches within drug databases to achieve a generalizable and efficient approach to drug repurposing. Our pipeline enables blind screening of any protein target without requiring prior structural or functional knowledge, allowing it to adapt to a wide range of diseases, including emerging health threats and novel targets where information is scarce. By rapidly generating potential ligands and efficiently identifying and ranking drug candidates, our approach accelerates the drug discovery process, broadening the scope and impact of repurposing efforts and offering new possibilities for therapeutic development. Detailed results and implementation can be accessed at https://github.com/HySonLab/DrugPipe.

bioinformatics↗

bioRxiv2024

DiscoEPG: A Python package for characterization of insect electrical penetration graph (EPG) signals

The Electrical Penetration Graph (EPG) technique is a well-known specialized tool that entomologists use to monitor and analyze the feeding behavior of piercing-sucking insects, such as aphids, whiteflies, and leafhoppers, on plants. Traditionally, the annotation is conducted by a well-trained technician who uses expert knowledge to compare the targeted waveforms with standard waveforms of aphid feeding behavior, which takes approximately 30 minutes to annotate an 8-hour recording depending on the complexity of the insect behaviors. Machine learning (ML) models, which shown great potential in monitoring insects behaviors, have recently been used to speed up this process. However, most publicly available tools that provide automatic annotation suffer from low prediction accuracy due to only using simple distinction rules to classify waveforms. For this reason, we develop DiscoEPG, an open-source Python package which performs accurate automatic EPG signal annotation. Various ML algorithms were experimented rigorously, which reports greater prediction power and improved accuracy in comparison to previous studies. In addition, we equipped our package with novel tools for generating journal-level plots to facilitate visual inspection, while including the computation of various EPG parameters and necessary statistical analysis which are popular in the research of aphids. With DiscoEPG, we aim to facilitate the rapid characterization of analysis of piercing-sucking insects feeding behavior through EPG signal, making this technique more viable to researchers who share the same interest. Our package is publicly available at: https://github.com/HySonLab/ML4Insects.

bioinformatics↗

bioRxiv2024

LightMed: A Light-weight and Robust FFT-Based Model for Adversarially Resilient Medical Image Segmentation

Accurate and reliable medical image segmentation is essential for computer-aided diagnosis and formulating appropriate treatment plans. However, real-world challenges such as suboptimal image quality and computational resource constraints hinder the effective deployment of deep learning-based segmentation models. To address these issues, we propose LightMed, a novel efficient neural architecture based on Fast Fourier Transform (FFT). Different from prior works, our model directly learns on the frequency domain, harnessing its resilience to noise and un-even brightness, which common artifacts found in medical images. By focusing on low-frequency image components, we significantly reduce computational complexity while preserving essential image features. Our deep learning architecture extracts discriminative features directly from the Fourier domain, leading to improved segmentation accuracy and robustness compared to traditional spatial domain methods. Additionally, we propose a new benchmark incorporating various levels of Gaussian noise to assess susceptibility to noise attacks. The experimental results demonstrate that LightMed not only effectively eliminates noise and consistently achieves accurate image segmentation but also shows robust resistance to imperceptible adversarial attacks compared to other baseline models. Our new benchmark datasets and source code are publicly available at https://github.com/HySonLab/LightMed.

bioinformatics↗

bioRxiv2024

VieMedKG: Knowledge Graph and Benchmark for Traditional Vietnamese Medicine

Traditional Vietnamese Medicine (TVM) and Traditional Chinese Medicine (TCM) have shared significant similarities due to their geographical location, cultural exchanges, and hot and humid climatic conditions. However, unlike TCM, which has substantial works published to construct a knowledge graph, there is a notable absence of a comprehensive knowledge graph for TVM. This paper presents the first endeavor to build a knowledge graph for TVM based on extensive existing resources from TCM. We name our knowledge graph as VieMedKG. We propose a translation and filtration process to adapt TCM knowledge graphs to TVM, identifying the overlapping and unique elements of TVM. In addition, the constructed knowledge graph is then exploited further for developing a curated benchmark for the knowledge graph-based question-answering problem with the potential to support doctors and patients in assisting doctors and patients in identifying various diseases. Our work will not only bridge the gap between TCM and TVM but also set the foundation for future research into traditional Vietnamese medicine community. Our source code is publicly available at https://github.com/HySonLab/VieMedKG/.

bioinformatics↗

bioRxiv2024

Machine learning for characterizing plant-insect interactions through electrical penetration graphic signal

The electrical penetration graph (EPG) is a well-known technique that provides insights into the feeding behavior of insects with piercing-sucking mouthparts, mostly hemipterans. Since its inception in the 1960s, EPG has become indispensable in studying plant-insect interactions, revealing critical information about host plant selection, plant resistance, virus transmission, and responses to environmental factors. By integrating the plant and insect into an electrical circuit, EPG allows researchers to identify specific feeding behaviors based on distinct waveform patterns associated with activities within plant tissues. However, the traditional manual analysis of EPG waveform data is time-consuming and labor-intensive, limiting research throughput. This study presents a novel machine-learning approach to automate the segmentation and classification of EPG signals. We rigorously evaluated six diverse machine learning models, including neural networks, tree-based models, and logistic regressions, using an extensive dataset from aphid feeding experiments. Our results demonstrate that a Residual Network (ResNet) architecture achieved the highest overall waveform classification accuracy of 96.8% and highest segmentation overlap rate of 84.4%, highlighting the potential of machine learning for accurate and efficient EPG analysis. This automated approach promises to accelerate research in this field significantly and has the potential to be generalized to other insect species and experimental settings. Our findings underscore the value of applying advanced computational techniques to complex biological datasets, paving the way for a more comprehensive understanding of insect-plant interactions and their broader ecological implications. The source code for all experiments conducted within this study is publicly available at https://github.com/HySonLab/ML4Insects. Author summaryInsect pests of the order Hemiptera pose a significant threat to global agriculture, causing substantial crop losses due to direct feeding and serving as vectors for many economically important plant viruses. Understanding plant-insect interactions is crucial for mitigating these impacts. The electrical penetration graph (EPG) is a valuable tool that provides detailed insights into these interactions. However, the analysis of EPG data is a time-consuming, labor-intensive process that can also be prone to operator errors. State-of-the-art machine learning (ML) algorithms can be trained to perform this task accurately and consistently. These advanced algorithms can automate identifying and classifying specific EPG waveform patterns associated with distinct insect feeding behaviors. Our machine learning models, trained on extensive aphid feeding data demonstrated high accuracy in classifying these waveforms, with Residual Network (ResNet) architecture achieving the best performance. The automated approach saves time and resources, eliminates operator error, and also enables the identification of novel feeding patterns, providing a deeper understanding of the mechanisms underlying plant-aphid interactions. Moreover, our evaluation of a large, diverse dataset of four aphid species on four host plants indicates the potential for generalizing these models to different experimental settings. By applying advanced computational techniques to EPG data, we are pioneering the intelligent surveillance of aphid feeding habits. This approach promises to significantly enhance our efforts in developing a better understanding of factors that affect aphid feeding.

ecology↗

bioRxiv2024

Complex-based Ligand-Binding Proteins Redesign by Equivariant Diffusion-based Generative Models

Proteins, serving as the fundamental architects of biological processes, interact with ligands to perform a myriad of functions essential for life. Designing functional ligand-binding proteins is pivotal for advancing drug development and enhancing therapeutic efficacy. In this study, we introduce ProteinReDiff, an efficient computational framework targeting the redesign of ligand-binding proteins. Using equivariant diffusion-based generative models, ProteinReDiff enables the creation of high-affinity ligand-binding proteins without the need for detailed structural information, leveraging instead the potential of initial protein sequences and ligand SMILES strings. Our evaluations across sequence diversity, structural preservation, and ligand binding affinity underscore ProteinReDiffs potential to advance computational drug discovery and protein engineering. Our source code is publicly available at https://github.com/HySonLab/Protein_Redesign.

bioinformatics↗

bioRxiv2024

Latent-based Directed Evolution accelerated by Gradient Ascent for Protein Sequence Design

Directed evolution has been the most effective method for protein engineering that optimizes biological functionalities through a resource-intensive process of screening or selecting among a vast range of mutations. To mitigate this extensive procedure, recent advancements in machine learning-guided methodologies center around the establishment of a surrogate sequence-function model. In this paper, we propose Latent-based Directed Evolution (LDE), an evolutionary algorithm designed to prioritize the exploration of high-fitness mutants in the latent space. At its core, LDE is a regularized variational autoencoder (VAE), harnessing the capabilities of the state-of-the-art Protein Language Model (pLM), ESM-2, to construct a meaningful latent space of sequences. From this encoded representation, we present a novel approach for efficient traversal on the fitness landscape, employing a combination of gradient-based methods and directed evolution. Experimental evaluations conducted on eight protein sequence design tasks demonstrate the superior performance of our proposed LDE over previous baseline algorithms. Our implementation is publicly available at https://github.com/HySonLab/LatentDE.

bioinformatics↗

bioRxiv2023

Multimodal Pretraining for Unsupervised Protein Representation Learning

In this paper, we introduce a framework of symmetry-preserving multimodal pretraining to learn a unified representation of proteins in an unsupervised manner, encompassing both primary and tertiary structures. Our approach involves proposing specific pretraining methods for sequences, graphs, and 3D point clouds associated with each protein structure, leveraging the power of large language models and generative models. We present a novel way to combining representations from multiple sources of information into a single global representation for proteins. We carefully analyze the performance of our framework in the pretraining tasks. For the fine-tuning tasks, our experiments have shown that our new multimodal representation can achieve competitive results in protein-ligand binding affinity prediction, protein fold classification, enzyme identification and mutation stability prediction. We expect that this work will accelerate future research in proteins. Our source code in PyTorch deep learning framework is publicly available at https://github.com/HySonLab/Protein_Pretrain.

bioinformatics↗

bioRxiv2023

Protein Design by Directed Evolution Guided by Large Language Models

Directed evolution, a strategy for protein engineering, optimizes protein properties (i.e., fitness) by a rigorous and resource-intensive process of screening or selecting among a vast range of mutations. By conducting an in silico screening of sequence properties, machine learning-guided directed evolution (MLDE) can expedite the optimization process and alleviate the experimental workload. In this work, we propose a general MLDE framework in which we apply recent advancements of Deep Learning in protein representation learning and protein property prediction to accelerate the searching and optimization processes. In particular, we introduce an optimization pipeline that utilizes Large Language Models (LLMs) to pinpoint the mutation hotspots in the sequence and then suggest replacements to improve the overall fitness. Our experiments have shown the superior efficiency and efficacy of our proposed framework in the conditional protein generation, in comparision with other state-of-the-art baseline algorithms. We expect this work will shed a new light on not only protein engineering but also on solving combinatorial problems using data-driven methods. Our implementation is publicly available at https://github.com/HySonLab/Directed_Evolution.

bioinformatics↗

bioRxiv2023

Target-aware Variational Auto-encoders for Ligand Generation with Multimodal Protein Representation Learning

Without knowledge of specific pockets, generating ligands based on the global structure of a protein target plays a crucial role in drug discovery as it helps reduce the search space for potential drug-like candidates in the pipeline. However, contemporary methods require optimizing tailored networks for each protein, which is arduous and costly. To address this issue, we introduce TargetVAE, a target-aware variational auto-encoder that generates ligands with high binding affinities to arbitrary protein targets, guided by a novel multimodal deep neural network built based on graph Transformers as the prior for the generative model. This is the first effort to unify different representations of proteins (e.g., sequence of amino-acids, 3D structure) into a single model that we name as Protein Multimodal Network (PMN). Our multimodal architecture learns from the entire protein structures and is able to capture their sequential, topological and geometrical information. We showcase the superiority of our approach by conducting extensive experiments and evaluations, including the assessment of generative model quality, ligand generation for unseen targets, docking score computation, and binding affinity prediction. Empirical results demonstrate the promising performance of our proposed approach. Our software package is publicly available at https://github.com/HySonLab/Ligand_Generation.

biochemistry↗