data.zip

IntroductionDeepRMSF is an automated deep-learning based approach for ‘imaging’ the dynamics of RNA at atomic resolution. Starting with a given PDB-formatted structure, DeepRMSF, it is first translated to density map, structure feature in density map is translated by structure. The maps are segmented into a series of density boxes, which served as input for the model. Finally, predicted RMSF subboxes were then merged into an RMSF map as the RMSF prediction map for this RNA. Source code for DeepRMSF: an automated computational method for RNA dynamics modeling by deep learningFilesrna_util.pyThis file undergoes the most basic data processing, such as generating simulated maps.rna_to_input.pyThis file divides the maps into boxes as input to the model.model_util.pyFunctions required for model training.rmsf_model.pyDeepRMSF model.rna_five_fold.pyWe train and test model using 5-fold cross-validation.main.pyUsageDownload the code file and run <b>main.py</b>. python main.py [-h] [--ori_dir ORI_DIR] [--data_dir DATA_DIR] [--box_file BOX_FILE] [--log_dir LOG_DIR]<br>You can enter the following parameters,--ori_dirThe folder for saving PDBs of RNAs.--data_dirThe folder where you want to save the simulated maps.--box_fileThe folder where you want to save the box_files.--log_dirThe folder where you want to save predicted data.-h or --helpYou can consult the help.InputPDBs of RNAs.OutputPDBs with predictive normalized RMSF values which replace B-factor values. The file names are "{pdbid}_pre_nor.pdb". These PDBs can be visualized by Chimera and so on.ExampleYou can view the <b>test</b> folder to learn about the DeepRMSF prediction process.Supporting softwaresx3DNA-DSSRTo obtain secondary structure.UCSF ChimeraSimulated maps were obtained with UCSF Chimera, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311.DeepRMSF<br>

# 引言
DeepRMSF是一种基于深度学习的自动化方法，可实现RNA动态特性的原子分辨率"成像"。当输入给定的PDB格式（PDB）结构文件时，DeepRMSF首先将其转换为密度图，再通过结构信息提取密度图中的结构特征。随后将这些密度图分割为一系列密度盒，作为模型的输入数据。最终，将预测得到的根均方波动（RMSF）子盒合并为RMSF图，即该RNA的RMSF预测图。

## DeepRMSF源代码：基于深度学习的RNA动态建模自动化计算方法
### 文件说明
1. `rna_util.py`：执行最基础的数据处理工作，例如生成模拟密度图。
2. `rna_to_input.py`：将密度图分割为盒状数据，作为模型的输入。
3. `model_util.py`：包含模型训练所需的各类工具函数。
4. `rmsf_model.py`：实现DeepRMSF模型主体架构。
5. `rna_five_fold.py`：通过5折交叉验证对模型进行训练与测试。
6. `main.py`：使用方法
获取代码文件后，可通过以下命令运行`main.py`：
bash
python main.py [-h] [--ori_dir ORI_DIR] [--data_dir DATA_DIR] [--box_file BOX_FILE] [--log_dir LOG_DIR]

可配置的参数说明如下：
- `--ori_dir`：存储RNA的PDB文件的目标文件夹路径。
- `--data_dir`：存储模拟密度图的目标文件夹路径。
- `--box_file`：存储盒状数据文件的目标文件夹路径。
- `--log_dir`：存储预测结果的目标文件夹路径。
- `-h`或`--help`：查看命令行帮助信息。

### 输入与输出
- **输入**：RNA的PDB格式结构文件。
- **输出**：替换了B因子值的、带有预测归一化RMSF值的PDB文件，文件命名格式为"{pdbid}_pre_nor.pdb"。此类PDB文件可通过Chimera（UCSF Chimera）等工具进行可视化。

### 示例
可查看`test`文件夹以了解DeepRMSF的完整预测流程。

### 配套软件
1. `x3DNA-DSSR`：用于获取RNA的二级结构信息。
2. `UCSF Chimera`：用于生成模拟密度图。UCSF Chimera由加州大学旧金山分校生物计算、可视化与信息学资源开发，受美国国立卫生研究院（NIH）P41-GM103311项目资助。
DeepRMSF