Wei Xingfei, Mo Qiankun, Chen Chi, Bathe Mark, Hernandez Rigoberto
Department of Chemistry, Johns Hopkins University, Baltimore, Maryland 21218, United States.
Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States.
J Chem Inf Model. 2025 Jul 14;65(13):6526-6536. doi: 10.1021/acs.jcim.5c00330. Epub 2025 Jun 20.
Artificial intelligence (AI) models remain an emerging strategy to accelerate materials design and development. We demonstrate that CNN models can characterize DNA origami nanostructures employed in programmable self-assembly, which is important in many applications such as in biomedicine. Specifically, we benchmark the performance of 9 CNN models, namely, AlexNet, GoogLeNet, VGG16, VGG19, ResNet18, ResNet34, ResNet50, ResNet101, and ResNet152 to characterize the ligation number of DNA origami nanostructures in transmission electron microscopy (TEM) images. We first pretrain CNN models using a large image data set of 720 images from our coarse-grained (CG) molecular dynamics (MD) simulations. Then, we fine-tune the pretrained CNN models, using a small experimental TEM data set with 146 TEM images. All CNN models were found to have similar computational time requirements, although their model sizes and performances are different. We use 20 test MD images to demonstrate that among all of the pretrained CNN models, ResNet50 and VGG16 have the highest and second-highest accuracies. Among the fine-tuned models, VGG16 was found to have the highest agreement with the test TEM images. Thus, we conclude that fine-tuned VGG16 models can quickly characterize the number of ligation sites of nanostructures in large TEM images.
人工智能（AI）模型仍然是加速材料设计与开发的一种新兴策略。我们证明了卷积神经网络（CNN）模型能够对可编程自组装中使用的DNA折纸纳米结构进行表征，这在生物医学等许多应用中都很重要。具体而言，我们对9种CNN模型的性能进行了基准测试，即AlexNet、GoogLeNet、VGG16、VGG19、ResNet18、ResNet34、ResNet50、ResNet101和ResNet152，以在透射电子显微镜（TEM）图像中表征DNA折纸纳米结构的连接数。我们首先使用来自粗粒度（CG）分子动力学（MD）模拟的720张图像的大型图像数据集对CNN模型进行预训练。然后，我们使用包含146张TEM图像的小型实验TEM数据集对预训练的CNN模型进行微调。尽管所有CNN模型的模型大小和性能不同，但发现它们具有相似的计算时间要求。我们使用20张测试MD图像来证明，在所有预训练的CNN模型中，ResNet50和VGG16具有最高和第二高的准确率。在微调模型中，发现VGG16与测试TEM图像的一致性最高。因此，我们得出结论，微调后的VGG16模型可以快速表征大型TEM图像中纳米结构的连接位点数量。
