分子如何影响细胞?揭开对比表观分子检索的神秘面纱

Philip Fradkin, Puria Azadi, Karush Suri, Frederik Wenkel, Ali Bashashati, Maciej Sypetkowski, Dominique Beaini
{"title":"分子如何影响细胞?揭开对比表观分子检索的神秘面纱","authors":"Philip Fradkin, Puria Azadi, Karush Suri, Frederik Wenkel, Ali Bashashati, Maciej Sypetkowski, Dominique Beaini","doi":"arxiv-2409.08302","DOIUrl":null,"url":null,"abstract":"Predicting molecular impact on cellular function is a core challenge in\ntherapeutic design. Phenomic experiments, designed to capture cellular\nmorphology, utilize microscopy based techniques and demonstrate a high\nthroughput solution for uncovering molecular impact on the cell. In this work,\nwe learn a joint latent space between molecular structures and microscopy\nphenomic experiments, aligning paired samples with contrastive learning.\nSpecifically, we study the problem ofContrastive PhenoMolecular Retrieval,\nwhich consists of zero-shot molecular structure identification conditioned on\nphenomic experiments. We assess challenges in multi-modal learning of phenomics\nand molecular modalities such as experimental batch effect, inactive molecule\nperturbations, and encoding perturbation concentration. We demonstrate improved\nmulti-modal learner retrieval through (1) a uni-modal pre-trained phenomics\nmodel, (2) a novel inter sample similarity aware loss, and (3) models\nconditioned on a representation of molecular concentration. Following this\nrecipe, we propose MolPhenix, a molecular phenomics model. MolPhenix leverages\na pre-trained phenomics model to demonstrate significant performance gains\nacross perturbation concentrations, molecular scaffolds, and activity\nthresholds. In particular, we demonstrate an 8.1x improvement in zero shot\nmolecular retrieval of active molecules over the previous state-of-the-art,\nreaching 77.33% in top-1% accuracy. These results open the door for machine\nlearning to be applied in virtual phenomics screening, which can significantly\nbenefit drug discovery applications.","PeriodicalId":501266,"journal":{"name":"arXiv - QuanBio - Quantitative Methods","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"How Molecules Impact Cells: Unlocking Contrastive PhenoMolecular Retrieval\",\"authors\":\"Philip Fradkin, Puria Azadi, Karush Suri, Frederik Wenkel, Ali Bashashati, Maciej Sypetkowski, Dominique Beaini\",\"doi\":\"arxiv-2409.08302\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Predicting molecular impact on cellular function is a core challenge in\\ntherapeutic design. Phenomic experiments, designed to capture cellular\\nmorphology, utilize microscopy based techniques and demonstrate a high\\nthroughput solution for uncovering molecular impact on the cell. In this work,\\nwe learn a joint latent space between molecular structures and microscopy\\nphenomic experiments, aligning paired samples with contrastive learning.\\nSpecifically, we study the problem ofContrastive PhenoMolecular Retrieval,\\nwhich consists of zero-shot molecular structure identification conditioned on\\nphenomic experiments. We assess challenges in multi-modal learning of phenomics\\nand molecular modalities such as experimental batch effect, inactive molecule\\nperturbations, and encoding perturbation concentration. We demonstrate improved\\nmulti-modal learner retrieval through (1) a uni-modal pre-trained phenomics\\nmodel, (2) a novel inter sample similarity aware loss, and (3) models\\nconditioned on a representation of molecular concentration. Following this\\nrecipe, we propose MolPhenix, a molecular phenomics model. MolPhenix leverages\\na pre-trained phenomics model to demonstrate significant performance gains\\nacross perturbation concentrations, molecular scaffolds, and activity\\nthresholds. In particular, we demonstrate an 8.1x improvement in zero shot\\nmolecular retrieval of active molecules over the previous state-of-the-art,\\nreaching 77.33% in top-1% accuracy. These results open the door for machine\\nlearning to be applied in virtual phenomics screening, which can significantly\\nbenefit drug discovery applications.\",\"PeriodicalId\":501266,\"journal\":{\"name\":\"arXiv - QuanBio - Quantitative Methods\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-10\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - QuanBio - Quantitative Methods\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.08302\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - QuanBio - Quantitative Methods","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.08302","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

预测分子对细胞功能的影响是治疗设计的核心挑战。表观实验旨在捕捉细胞形态,利用基于显微镜的技术,展示了揭示分子对细胞影响的高通量解决方案。在这项工作中,我们学习分子结构和显微镜表观实验之间的联合潜空间,通过对比学习对配对样本进行对齐。具体来说,我们研究了对比表观分子检索(Contrastive PhenoMolecular Retrieval)问题,该问题包括以表观实验为条件的零次分子结构识别。我们评估了表型组学和分子模式的多模式学习所面临的挑战,如实验批次效应、非活性分子扰动和编码扰动浓度。我们展示了通过(1)单模态预训练表型组学模型,(2)新颖的样本间相似性感知损失,以及(3)以分子浓度表示为条件的模型,改进了多模态学习器检索。根据这一思路,我们提出了分子表型组学模型 MolPhenix。MolPhenix 利用预先训练好的表型组学模型,在各种扰动浓度、分子支架和活动阈值上都取得了显著的性能提升。特别是在活性分子的零镜头分子检索方面,我们比以前的先进水平提高了 8.1 倍,最高准确率达到 77.33%。这些结果为机器学习在虚拟表型组学筛选中的应用打开了大门,可极大地促进药物发现应用。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
How Molecules Impact Cells: Unlocking Contrastive PhenoMolecular Retrieval
Predicting molecular impact on cellular function is a core challenge in therapeutic design. Phenomic experiments, designed to capture cellular morphology, utilize microscopy based techniques and demonstrate a high throughput solution for uncovering molecular impact on the cell. In this work, we learn a joint latent space between molecular structures and microscopy phenomic experiments, aligning paired samples with contrastive learning. Specifically, we study the problem ofContrastive PhenoMolecular Retrieval, which consists of zero-shot molecular structure identification conditioned on phenomic experiments. We assess challenges in multi-modal learning of phenomics and molecular modalities such as experimental batch effect, inactive molecule perturbations, and encoding perturbation concentration. We demonstrate improved multi-modal learner retrieval through (1) a uni-modal pre-trained phenomics model, (2) a novel inter sample similarity aware loss, and (3) models conditioned on a representation of molecular concentration. Following this recipe, we propose MolPhenix, a molecular phenomics model. MolPhenix leverages a pre-trained phenomics model to demonstrate significant performance gains across perturbation concentrations, molecular scaffolds, and activity thresholds. In particular, we demonstrate an 8.1x improvement in zero shot molecular retrieval of active molecules over the previous state-of-the-art, reaching 77.33% in top-1% accuracy. These results open the door for machine learning to be applied in virtual phenomics screening, which can significantly benefit drug discovery applications.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信