arXiv - QuanBio - Biomolecules最新文献

{"title":"Design of Ligand-Binding Proteins with Atomic Flow Matching","authors":"Junqi Liu, Shaoning Li, Chence Shi, Zhi Yang, Jian Tang","doi":"arxiv-2409.12080","DOIUrl":"https://doi.org/arxiv-2409.12080","url":null,"abstract":"Designing novel proteins that bind to small molecules is a long-standing\u0000challenge in computational biology, with applications in developing catalysts,\u0000biosensors, and more. Current computational methods rely on the assumption that\u0000the binding pose of the target molecule is known, which is not always feasible,\u0000as conformations of novel targets are often unknown and tend to change upon\u0000binding. In this work, we formulate proteins and molecules as unified\u0000biotokens, and present AtomFlow, a novel deep generative model under the\u0000flow-matching framework for the design of ligand-binding proteins from the 2D\u0000target molecular graph alone. Operating on representative atoms of biotokens,\u0000AtomFlow captures the flexibility of ligands and generates ligand conformations\u0000and protein backbone structures iteratively. We consider the multi-scale nature\u0000of biotokens and demonstrate that AtomFlow can be effectively trained on a\u0000subset of structures from the Protein Data Bank, by matching flow vector field\u0000using an SE(3) equivariant structure prediction network. Experimental results\u0000show that our method can generate high fidelity ligand-binding proteins and\u0000achieve performance comparable to the state-of-the-art model RFDiffusionAA,\u0000while not requiring bound ligand structures. As a general framework, AtomFlow\u0000holds the potential to be applied to various biomolecule generation tasks in\u0000the future.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"40 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142252789","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"How to explain the sensitivity of DNA double-strand breaks yield to 125I position?","authors":"Mario Enrique Alcocer ÁvilaCELIA, IP2I Lyon, Elif HindiéIUF, INCIA, Christophe ChampionCELIA","doi":"arxiv-2409.11185","DOIUrl":"https://doi.org/arxiv-2409.11185","url":null,"abstract":"Purpose: Auger emitters exhibit interesting features due to their emission of\u0000a cascade of short-range Auger electrons. Maximum DNA breakage efficacy is\u0000achieved when decays occur near DNA. Studies of double-strand breaks (DSBs)\u0000yields in plasmids revealed cutoff distances from DNA axis of 10.5-12 {AA},\u0000beyond which the mechanism of DSBs moves from direct to indirect effects, and\u0000the yield decreases rapidly. Some authors suggested that the average energy\u0000deposited in a DNA cylinder could explain such cutoffs. We aimed to study this\u0000hypothesis in further detail.Materials and methods: Using the Monte Carlo code\u0000CELLDOSE, we investigated the influence of the 125I atom position on energy\u0000deposits and absorbed doses per decay not only in a DNA cylinder, but also in\u0000individual strands, each modeled as 10 spheres encompassing the fragility sites\u0000for phosphodiester bond cleavage.Results: The dose per decay decreased much\u0000more rapidly for a sphere in the proximal strand than for the DNA cylinder. For\u0000example, when moving the 125I source from 10.5 {AA} to 11.5 {AA}, the average\u0000dose to the sphere dropped by 43%, compared to only 13% in the case of the\u0000cylinder.Conclusions: Explaining variations in DSBs yields with 125I position\u0000should consider the probability of inducing damage in the proximal strand\u0000(nearest to the 125I atom). The energy received by fragility sites in this\u0000strand is highly influenced by the isotropic (4$pi$) emission of 125I\u0000low-energy Auger electrons. The positioning of Auger emitters for targeted\u0000radionuclide therapy can be envisioned accordingly.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"175 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142252790","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"GFlowNet Pretraining with Inexpensive Rewards","authors":"Mohit Pandey, Gopeshh Subbaraj, Emmanuel Bengio","doi":"arxiv-2409.09702","DOIUrl":"https://doi.org/arxiv-2409.09702","url":null,"abstract":"Generative Flow Networks (GFlowNets), a class of generative models have\u0000recently emerged as a suitable framework for generating diverse and\u0000high-quality molecular structures by learning from unnormalized reward\u0000distributions. Previous works in this direction often restrict exploration by\u0000using predefined molecular fragments as building blocks, limiting the chemical\u0000space that can be accessed. In this work, we introduce Atomic GFlowNets\u0000(A-GFNs), a foundational generative model leveraging individual atoms as\u0000building blocks to explore drug-like chemical space more comprehensively. We\u0000propose an unsupervised pre-training approach using offline drug-like molecule\u0000datasets, which conditions A-GFNs on inexpensive yet informative molecular\u0000descriptors such as drug-likeliness, topological polar surface area, and\u0000synthetic accessibility scores. These properties serve as proxy rewards,\u0000guiding A-GFNs towards regions of chemical space that exhibit desirable\u0000pharmacological properties. We further our method by implementing a\u0000goal-conditioned fine-tuning process, which adapts A-GFNs to optimize for\u0000specific target properties. In this work, we pretrain A-GFN on the ZINC15\u0000offline dataset and employ robust evaluation metrics to show the effectiveness\u0000of our approach when compared to other relevant baseline methods in drug\u0000design.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"65 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142252791","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"One-way catalysis: Insights from a solvable lattice model","authors":"Sara Mahdavi, Yann Sakref, Olivier Rivoire","doi":"arxiv-2409.09421","DOIUrl":"https://doi.org/arxiv-2409.09421","url":null,"abstract":"Catalysts speed up chemical reactions with no energy input and without being\u0000transformed in the process, therefore leaving equilibrium constants unchanged.\u0000Some catalysts, however, are much more efficient at accelerating one direction\u0000of a reaction. How is such an asymmetry consistent with chemical equilibrium,\u0000where as many forward and reverse reactions must occur? We use the rigorous\u0000framework of a simple but exactly solvable lattice model to study this question\u0000in the context of a strictly one-way catalyst, which can only accelerate one\u0000way of a reaction. The model illustrates a mathematical relationship between\u0000asymmetric transition rates, which underlie directional catalysis, and\u0000symmetric transition fluxes, which underlie chemical equilibrium. The degree of\u0000directionality generally depends on the catalytic mechanism and we compare\u0000different mechanisms to show how they can obey different scaling laws. The\u0000results showcase the ability of simple physical models to address fundamental\u0000questions in catalysis.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"10 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142252792","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Diffusion and Spectroscopy of H$_2$ in Myoglobin","authors":"Jiri Käser, Kai Töpfer, Markus Meuwly","doi":"arxiv-2409.08737","DOIUrl":"https://doi.org/arxiv-2409.08737","url":null,"abstract":"The diffusional dynamics and vibrational spectroscopy of molecular hydrogen\u0000(H$_2$) in myoglobin (Mb) is characterized. Hydrogen has been implicated in a\u0000number of physiologically relevant processes, including cellular aging or\u0000inflammation. Here, the internal diffusion through the protein matrix was\u0000characterized and the vibrational spectroscopy was investigated using\u0000conventional empirical energy functions and improved models able to describe\u0000higher-order electrostatic moments of the ligand. H$_2$ can occupy the same\u0000internal defects as already found for Xe or CO (Xe1 to Xe4 and B-state).\u0000Furthermore, 4 additional sites were found, some of which had been discovered\u0000in earlier simulation studies. The vibrational spectra using the most refined\u0000energy function indicate that depending on the docking site the spectroscopy of\u0000H$_2$ differs. The maxima of the absorption spectra cover $sim 20$ cm$^{-1}$\u0000which are indicative of a pronounced effect of the surrounding protein matrix\u0000on the vibrational spectroscopy of the ligand. Electronic structure\u0000calculations show that H$_2$ forms a stable complex with the heme-iron\u0000(stabilized by $sim -12$ kcal/mol) but splitting of H$_2$ is unlikely due to a\u0000high activation energy ($sim 50$ kcal/mol).","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"116 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142252795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mechanosensitive ion channels: Old but new","authors":"Uğur Çetiner","doi":"arxiv-2409.09200","DOIUrl":"https://doi.org/arxiv-2409.09200","url":null,"abstract":"Ion channels orchestrate the communication between cells and their\u0000environment. These are special proteins capable of changing their shape to\u0000allow ions to pass through membranes in response to stimuli like membrane\u0000tension. As the letters are written and exchanged in the form of ionic\u0000currents, the energetic cost of this communication is of interest to many. To\u0000this end, we introduce the nanoscale thermodynamics of mechanosensitive ion\u0000channels, which amounts to defining work and heat for traces obtained during\u0000patch-clamp experiments. We also discuss the interplay between ion channel\u0000physics and evolution by showing how information-processing capabilities are\u0000coupled with the energy landscapes of channels.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"5 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142252794","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Quantum-inspired Reinforcement Learning for Synthesizable Drug Design","authors":"Dannong Wang, Jintai Chen, Zhiding Liang, Tianfan Fu, Xiao-Yang Liu","doi":"arxiv-2409.09183","DOIUrl":"https://doi.org/arxiv-2409.09183","url":null,"abstract":"Synthesizable molecular design (also known as synthesizable molecular\u0000optimization) is a fundamental problem in drug discovery, and involves\u0000designing novel molecular structures to improve their properties according to\u0000drug-relevant oracle functions (i.e., objective) while ensuring synthetic\u0000feasibility. However, existing methods are mostly based on random search. To\u0000address this issue, in this paper, we introduce a novel approach using the\u0000reinforcement learning method with quantum-inspired simulated annealing policy\u0000neural network to navigate the vast discrete space of chemical structures\u0000intelligently. Specifically, we employ a deterministic REINFORCE algorithm\u0000using policy neural networks to output transitional probability to guide state\u0000transitions and local search using genetic algorithm to refine solutions to a\u0000local optimum within each iteration. Our methods are evaluated with the\u0000Practical Molecular Optimization (PMO) benchmark framework with a 10K query\u0000budget. We further showcase the competitive performance of our method by\u0000comparing it against the state-of-the-art genetic algorithms-based method.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"20 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142252793","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"De novo design of high-affinity protein binders with AlphaProteo","authors":"Vinicius Zambaldi, David La, Alexander E. Chu, Harshnira Patani, Amy E. Danson, Tristan O. C. Kwan, Thomas Frerix, Rosalia G. Schneider, David Saxton, Ashok Thillaisundaram, Zachary Wu, Isabel Moraes, Oskar Lange, Eliseo Papa, Gabriella Stanton, Victor Martin, Sukhdeep Singh, Lai H. Wong, Russ Bates, Simon A. Kohl, Josh Abramson, Andrew W. Senior, Yilmaz Alguel, Mary Y. Wu, Irene M. Aspalter, Katie Bentley, David L. V. Bauer, Peter Cherepanov, Demis Hassabis, Pushmeet Kohli, Rob Fergus, Jue Wang","doi":"arxiv-2409.08022","DOIUrl":"https://doi.org/arxiv-2409.08022","url":null,"abstract":"Computational design of protein-binding proteins is a fundamental capability\u0000with broad utility in biomedical research and biotechnology. Recent methods\u0000have made strides against some target proteins, but on-demand creation of\u0000high-affinity binders without multiple rounds of experimental testing remains\u0000an unsolved challenge. This technical report introduces AlphaProteo, a family\u0000of machine learning models for protein design, and details its performance on\u0000the de novo binder design problem. With AlphaProteo, we achieve 3- to 300-fold\u0000better binding affinities and higher experimental success rates than the best\u0000existing methods on seven target proteins. Our results suggest that AlphaProteo\u0000can generate binders \"ready-to-use\" for many research applications using only\u0000one round of medium-throughput screening and no further optimization.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"72 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142226968","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Perspective on AI-Guided Molecular Simulations in VR: Exploring Strategies for Imitation Learning in Hyperdimensional Molecular Systems","authors":"Mohamed Dhouioui, Jonathan Barnoud, Rhoslyn Roebuck Williams, Harry J. Stroud, Phil Bates, David R. Glowacki","doi":"arxiv-2409.07189","DOIUrl":"https://doi.org/arxiv-2409.07189","url":null,"abstract":"Molecular dynamics simulations are a crucial computational tool for\u0000researchers to understand and engineer molecular structure and function in\u0000areas such as drug discovery, protein engineering, and material design. Despite\u0000their utility, MD simulations are expensive, owing to the high dimensionality\u0000of molecular systems. Interactive molecular dynamics in virtual reality\u0000(iMD-VR) has recently been developed as a 'human-in-the-loop' strategy, which\u0000leverages high-performance computing to accelerate the researcher's ability to\u0000solve the hyperdimensional sampling problem. By providing an immersive 3D\u0000environment that enables visualization and manipulation of real-time molecular\u0000motion, iMD-VR enables researchers and students to efficiently and intuitively\u0000explore and navigate these complex, high-dimensional systems. iMD-VR platforms\u0000offer a unique opportunity to quickly generate rich datasets that capture human\u0000experts' spatial insight regarding molecular structure and function. This paper\u0000explores the possibility of employing user-generated iMD-VR datasets to train\u0000AI agents via imitation learning (IL). IL is an important technique in robotics\u0000that enables agents to mimic complex behaviors from expert demonstrations, thus\u0000circumventing the need for explicit programming or intricate reward design. We\u0000review the utilization of IL for manipulation tasks in robotics and discuss how\u0000iMD-VR recordings could be used to train IL models for solving specific\u0000molecular 'tasks'. We then investigate how such approaches could be applied to\u0000the data captured from iMD-VR recordings. Finally, we outline the future\u0000research directions and potential challenges of using AI agents to augment\u0000human expertise to efficiently navigate conformational spaces, highlighting how\u0000this approach could provide valuable insight across domains such as materials\u0000science, protein engineering, and computer-aided drug design.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"73 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142216233","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AbGPT: De Novo Antibody Design via Generative Language Modeling","authors":"Desmond Kuan, Amir Barati Farimani","doi":"arxiv-2409.06090","DOIUrl":"https://doi.org/arxiv-2409.06090","url":null,"abstract":"The adaptive immune response, largely mediated by B-cell receptors (BCRs),\u0000plays a crucial role for effective pathogen neutralization due to its diversity\u0000and antigen specificity. Designing BCRs de novo, or from scratch, has been\u0000challenging because of their complex structure and diverse binding\u0000requirements. Protein language models (PLMs) have shown remarkable performance\u0000in contextualizing and performing various downstream tasks without relying on\u0000structural information. However, these models often lack a comprehensive\u0000understanding of the entire protein space, which limits their application in\u0000antibody design. In this study, we introduce Antibody Generative Pretrained\u0000Transformer (AbGPT), a model fine-tuned from a foundational PLM to enable a\u0000more informed design of BCR sequences. Using a custom generation and filtering\u0000pipeline, AbGPT successfully generated a high-quality library of 15,000 BCR\u0000sequences, demonstrating a strong understanding of the intrinsic variability\u0000and conserved regions within the antibody repertoire.","PeriodicalId":501022,"journal":{"name":"arXiv - QuanBio - Biomolecules","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142216234","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}