请根据需要预测的分子形式，选用对应的模块或流程，如图所示：

比较常用的模块推荐如下：

以下是部分功能模块的说明：

教程

结构预测介绍文档

Please select the corresponding module or workflow based on the molecular format you need to predict, as shown in the figure:

The most commonly used module recommendations are as follows:

The following are descriptions of some functional modules:

Tutorial

Structure Prediction Introduction Document

对PDB结构文件进行处理，包括去除杂质、补全确实原子或残基、加氢、修改链名或残基编号等。

教程

模拟结构处理介绍文档

Processing of PDB structure files includes removing impurities, completing missing atoms or residues, adding hydrogen atoms, modifying chain names or residue numbers, etc.

Tutorial

Introduction to Simulated Structure Processing Documentation

RMSD (Root Mean Square Deviation) 和 DockQ 都是评估分子结构相似性和对接模型质量的指标，但它们的应用范围和考量因素有所不同。

RMSD (Root Mean Square Deviation)

定义： RMSD 是衡量两个叠加的分子结构之间原子位置平均偏差的量度。它通过计算对应原子（通常是主链原子，如 Cα 原子，或所有重原子）在三维空间中的距离平方的均值，再开平方根得到。

应用场景：

• 蛋白质结构比较： 广泛用于比较两个或多个蛋白质结构之间的相似性，例如评估预测结构与实验结构的一致性，或者比较不同构象下的蛋白质结构。
• 分子动力学模拟： 用于监测模拟过程中分子结构随时间的稳定性，计算结构相对于初始构象的偏差。
• 小分子对接： 在小分子与蛋白质对接中，可以用来衡量预测的配体结合姿态与已知实验姿态的相似性。

特点：

• 单位是距离（Ångström）： RMSD 值越小，表示结构越相似。
• 对全局结构敏感： 即使少量原子的大偏差也会导致 RMSD 值显著增加。
• 不区分界面和非界面区域： RMSD 是对整个结构或指定原子集的整体比较，不特别关注分子间的相互作用界面。

DockQ

定义： DockQ 是一个专门用于评估蛋白质-蛋白质对接模型质量的连续性指标，范围在 [0,1] 之间。它结合了多个衡量对接质量的关键因素，以提供一个更全面、更接近 CAPRI (Critical Assessment of PRediction of Interactions) 评估标准的单一分数。

组成部分： DockQ 综合了以下几个关键指标：

• 界面 RMSD (iRMSD)： 衡量预测界面原子与真实界面原子之间的 RMSD。它只关注相互作用区域的结构偏差。
• 配体 RMSD (LRMSD)： 衡量配体分子（例如，对接中的一个小蛋白质）与真实配体位置的 RMSD。
• 天然接触分数 (Fnat)： 衡量预测模型中与真实复合物中相同接触点的比例。

计算方式： DockQ 并非简单的线性组合，而是通过对这些组分进行非线性变换和组合得出的，旨在更好地重现 CAPRI 的质量分类（Incorrect, Acceptable, Medium, High）。

应用场景：

• 蛋白质-蛋白质对接： 主要用于评估蛋白质-蛋白质对接预测模型的质量，判断预测的结合姿态是否准确。
• 对接方法开发和比较： 作为标准度量，用于评估不同对接算法的性能。

特点：

• 连续分数（0-1）： DockQ 值越高，表示对接模型质量越好（0代表质量差，1代表完美）。
• 关注界面质量： DockQ 特别强调界面区域的准确性，这对于评估蛋白质-蛋白质相互作用至关重要。
• 综合性指标： 它不仅仅是几何上的相似性，还考虑了接触点的正确性，因此比单纯的 RMSD 更能反映对接的生物学意义。
• 与 CAPRI 评估相关： DockQ 的设计初衷是为了更好地反映 CAPRI 比赛中使用的复杂评估标准，从而实现对接模型质量的标准化和可解释性比较。

RMSD 和 DockQ 的主要区别总结

简而言之，RMSD 是一个更通用的几何相似性度量，可以用于各种分子结构比较。而 DockQ 则是一个专门为蛋白质-蛋白质对接模型设计的高度集成的质量评估指标，它更全面地反映了对接的生物学相关性和准确性，因为它综合了界面几何精度和关键相互作用的正确性。在评估蛋白质-蛋白质对接时，DockQ 通常被认为是更优选和更具代表性的指标。

RMSD (Root Mean Square Deviation) and DockQ are both metrics used to evaluate molecular structure similarity and docking model quality, but they differ in their range of applications and considerations.

RMSD (Root Mean Square Deviation)

Definition: RMSD is a measure of the average deviation in atomic positions between two superimposed molecular structures. It is calculated by taking the square root of the mean of the squared distances between corresponding atoms (typically backbone atoms, such as Cα atoms, or all heavy atoms) in three-dimensional space.

Applications:

• Protein Structure Comparison: Widely used to compare the similarity between two or more protein structures, such as assessing the consistency between predicted and experimental structures, or comparing protein structures in different conformations.
• Molecular Dynamics Simulations: Used to monitor the stability of molecular structures over time during simulations, calculating deviations relative to the initial conformation.
• Small Molecule Docking: In small molecule-protein docking, it can be used to assess the similarity between predicted ligand binding poses and known experimental poses.

Characteristics:

• Measured in Distance (Ångström): The smaller the RMSD value, the more similar the structures are.
• Sensitive to Global Structure: Even a small number of atoms with large deviations can significantly increase the RMSD value.
• Does Not Distinguish Interface and Non-interface Regions: RMSD is a global comparison of the entire structure or a specified set of atoms, without special focus on interaction interfaces.

DockQ

Definition: DockQ is a continuous metric specifically designed to evaluate the quality of protein-protein docking models, ranging from [0,1]. It combines multiple key factors for assessing docking quality to provide a more comprehensive score that aligns closely with CAPRI (Critical Assessment of PRediction of Interactions) evaluation standards.

Components: DockQ integrates the following key metrics:

• Interface RMSD (iRMSD): Measures the RMSD between predicted interface atoms and true interface atoms, focusing only on structural deviations in the interaction region.
• Ligand RMSD (LRMSD): Measures the RMSD of the ligand molecule (e.g., a small protein in the docking) relative to its true position.
• Fraction of Native Contacts (Fnat): Measures the proportion of contact points in the predicted model that are the same as those in the true complex.

Calculation Method: DockQ is not a simple linear combination but is derived through nonlinear transformations and combinations of these components, aiming to better reproduce CAPRI’s quality classifications (Incorrect, Acceptable, Medium, High).

Applications:

• Protein-Protein Docking: Mainly used to assess the quality of protein-protein docking prediction models, determining whether the predicted binding poses are accurate.
• Docking Method Development and Comparison: Used as a standard measure to evaluate the performance of different docking algorithms.

Characteristics:

• Continuous Score (0-1): The higher the DockQ value, the better the quality of the docking model (0 indicates poor quality, 1 indicates perfect quality).
• Focus on Interface Quality: DockQ emphasizes the accuracy of the interface region, which is crucial for evaluating protein-protein interactions.
• Comprehensive Metric: It considers not only geometric similarity but also the correctness of contact points, making it more reflective of the biological significance of docking compared to RMSD alone.
• Related to CAPRI Evaluation: DockQ is designed to better reflect the complex evaluation standards used in the CAPRI competition, enabling standardized and interpretable comparisons of docking model quality.

Summary of Main Differences Between RMSD and DockQ

In short, RMSD is a more general metric for geometric similarity, applicable to various molecular structure comparisons. DockQ, on the other hand, is a highly integrated quality assessment metric specifically designed for protein-protein docking models, providing a more comprehensive reflection of the biological relevance and accuracy of docking by integrating interface geometric precision and the correctness of key interactions. In evaluating protein-protein docking, DockQ is often considered a more preferred and representative metric.

对于入门用户，可使用

• 全自动抗体人源化流程Antibody Humanization
• 纳米抗体人源化流程Nanobody Humanization

人源化流程介绍文档

对于有经验的用户，建议使用WeSeq中的交互式人源化设计面板：WeSeq->Humanization，可以基于人工经验调整人源化设计方案，实时对照结构进行更精细化的设计。

人源化-WeSeq介绍文档

For beginner users, the following can be used:

• Fully Automatic Antibody Humanization Process Antibody Humanization
• Nanobody Humanization Process Nanobody Humanization

Humanization Process Introduction Document

For experienced users, it is recommended to use the interactive humanization design panel in WeSeq: WeSeq->Humanization, which allows for manual adjustment of the humanization design scheme based on experience and real-time comparison with the structure for more refined design.

Humanization-WeSeq Introduction Document

免疫原性预测已经历多个版本迭代，目前推荐版本为：WeADApt v4.3，WeADApt v4.2, AlphaMHC v3.0 beta。

同时也可以从WeSeq中提交预测：WeSeq->Immunogenicity，界面更友好（推荐v4）。

免疫原性介绍文档

Immunogenicity prediction has undergone multiple version iterations. The currently recommended versions are: WeADApt v4.3, WeADApt v4.2, AlphaMHC v3.0 beta.

You can also submit predictions from WeSeq: WeSeq->Immunogenicity, which offers a more user-friendly interface (v4 recommended).

Immunogenicity Introduction Documentation

• 基于复合物结构的方法
  针对复合物结构的相互作用界面进行饱和突变或进化优势突变，再用物理方法计算能量。建议使用亲和力成熟流程：

  • 抗体亲和力成熟，使用Antibody Virtual Affinity Maturation流程。

  • 蛋白亲和力成熟，使用Protein Virtual Affinity Maturation流程。

    亲和力成熟介绍文档

• 基于配体的方法
  建议使用模块AA Probability Prediction，大语言模型预测高概率AA。主要是利用大语言模型和配体的序列（或结构）直接推荐高适应性（fitness）突变。

• 蛋白复合物亲和力的相对结合自由能计算，可使用模块Protein FEP。

• 基于ESMIF逆折叠模型，预测能提升结构亲和力的单点或多点突变，可使用模块Structure Evolution。

Methods Based on Complex Structure

For interaction interfaces of complex structures, perform saturation mutations or evolutionarily advantageous mutations, then use physical methods to calculate energy. It is recommended to use the affinity maturation workflows:

• For antibody affinity maturation, use Antibody Virtual Affinity Maturation.

• For protein affinity maturation, use Protein Virtual Affinity Maturation.

  Affinity Maturation Introduction Document

Methods Based on Ligands

• Ligand-based methods It is recommended to use the module AA Probability Prediction, where large language models predict high-probability amino acids. This approach primarily leverages large language models along with the ligand’s sequence (or structure) to directly recommend high-fitness mutations.

• For calculating the relative binding free energy of protein complex affinity, the module Protein FEP can be used.

• Based on the ESM-IF inverse folding model, to predict single or multiple mutations that can enhance structural affinity, the module Structure Evolution can be used.

热稳定性与蛋白的折叠自由能正相关，可能影响表达、纯度、PK等，优化方式包括基于物理的能量计算和ML/AI模型。

• 优化抗体稳定性，可使用最新版Antibody Stability Optimization。
  抗体稳定性优化流程介绍文档

• 优化蛋白稳定性，可使用最新版Protein Stability Optimization。
  蛋白稳定性优化流程介绍文档

• 预测蛋白质的绝对稳定性，可使用Absolute Folding Stability。
  蛋白绝对稳定性预测介绍文档

• 预测蛋白稳定性相对结合自由能，可使用Protein FEP。

• 基于ThermoMPNN模型预测蛋白质单点突变的稳定性变化，可使用Mutation Energy of Stability (ThermoMPNN)。

• 基于序列预测蛋白中潜在的PTM位点，可使用PTM Hotspot by Sequence。建议在WeSeq中进行分析：WeSeq->PTM。

  

• 基于结构预测蛋白中潜在的PTM位点，可使用PTM Hotspot by Structure。

• 基于ESMIF逆折叠模型，预测能提升结构稳定性的单点或多点突变，可使用Structure Evolution。

Thermal stability is positively correlated with the folding free energy of proteins, which may affect expression, purity, pharmacokinetics (PK), etc. Optimization methods include physics-based energy calculations and ML/AI models.

• To optimize antibody stability, you can use Antibody Stability Optimization.
  Antibody Stability Optimization Process Introduction Document

• To optimize protein stability, you can use Protein Stability Optimization.
  Protein Stability Optimization Process Introduction Document

• To predict the absolute stability of proteins, you can use Absolute Folding Stability.
  Absolute Folding Stability Prediction Introduction Document

• To predict the relative binding free energy of protein stability, you can use Protein FEP.

• To predict the stability changes of protein single-point mutations based on the ThermoMPNN model, you can use Mutation Energy of Stability (ThermoMPNN).

• To predict potential PTM sites in proteins based on sequence, you can use PTM Hotspot by Sequence. It is recommended to perform the analysis in WeSeq: WeSeq -> PTM.

  

• To predict potential PTM sites in proteins based on structure, you can use PTM Hotspot by Structure.

可开发性包括蛋白表面patch分析、理化性质计算（含pI）、TAP原则、PTM（基于序列）、基于结构的异构化预测、断裂位点预测等。

• 成药性一键综合评价

  • 抗体进行成药性一键综合评价，可以使用抗体可开发性预测流程：Antibody Developability Properties。
  • 纳米抗体成药性一键综合评价，可使用纳米抗体可开发性预测流程：Nanobody Developability Properties。
  • 同时，也可以在WeSeq中进行抗体可开发性预测分析，WeSeq->Developability->Antibody General Evaluation/Nanobody General Evaluation。

  
  

• 抗体QC流程

  • 一键计算抗体的人源性、成药性、免疫原性、脱靶等抗体所有性质可使用Antibody QC流程。
  • 同时，也可以在WeSeq中进行抗体QC预测，WeSeq->Developability->Antibody QC。

  

• Patch分析

  • 建议从WeView中运行：WeView->Analysis->Patch。Patch分析介绍文档

    

• PTM预测

  • 基于序列的PTM预测，建议直接在WeSeq运行：WeSeq->PTM。PTM预测介绍文档

    

  • 基于结构的PTM预测，可以直接在模块中运行：PTM Hotspot by Structure。

• 抗体成药性预测（TAP）

  • 可以直接在模块中运行：Therapeutic Antibody Profiler。TAP介绍文档

• 溶解度预测

  • 可以直接在模块中运行：Solubility Score或Solubility Score (CamSol)。溶解度预测介绍文档

• 聚集度预测

  • 可以直接在Aggregation Score模块中运行。聚集度预测介绍文档

Developability includes protein surface patch analysis, physicochemical property calculations (including pI), TAP principles, PTM (sequence-based), structure-based isomerization prediction, cleavage site prediction, etc.

• One-click Comprehensive Druggability Assessment

  • For antibody druggability comprehensive assessment, you can use the antibody developability prediction workflow: Antibody Developability Properties.
  • For nanobody druggability comprehensive assessment, you can use the nanobody developability prediction workflow: Nanobody Developability Properties
  • Additionally, antibody developability prediction analysis can also be performed in WeSeq: WeSeq->Developability->Antibody General Evaluation/Nanobody General Evaluation.

  
  

• Antibody QC Workflow

  • One-click calculation of all antibody properties including humanness, druggability, immunogenicity, off-target effects, etc. Antibody QC
  • Additionally, antibody QC prediction can also be performed in WeSeq: WeSeq->Developability->Antibody QC.

  

• Patch Analysis

  • Recommended to run from WeView: WeView->Analysis->Patch. Patch Analysis Documentation

    

• PTM Prediction

  • For sequence-based PTM prediction, it is recommended to run directly in WeSeq: WeSeq->PTM. PTM Prediction Documentation

    

  • For structure-based PTM prediction, you can run directly in the module: PTM Hotspot by Structure.

• Antibody Druggability Prediction (TAP)

  • Can be run directly in the module: Therapeutic Antibody Profiler. TAP Documentation

• Solubility Prediction

  • Can be run directly in modules: Solubility Score, Solubility Score (CamSol). Solubility Prediction Documentation

• Aggregation Prediction

  • Can be run directly in the module: Aggregation Score. Aggregation Prediction Documentation

序列分析包括序列编号、多序列比对、测序数据分析、频率分析、序列突变等。

• 序列编号
  进行抗体序列编号，建议在WeSeq中运行：WeSeq->Number。序列编号介绍文档

  

• 多序列比对
  进行多序列比对，建议在WeSeq中运行：WeSeq->Align。多序列比对介绍文档

  

• 测序数据分析
  进行测序数据分析，可以使用模块NGS Analysis。NGS Analysis介绍文档

• 频率分析
  进行频率分析，建议在WeSeq运行，WeSeq->Frequency。频率分析介绍文档

  

• 序列突变
  进行序列突变，建议在WeSeq中操作：WeSeq->Edit->Batch Mutate。或者使用Sequence Mutation模块。

  

Sequence analysis includes sequence numbering, multiple sequence alignment, sequencing data analysis, frequency analysis, and sequence mutations.

Sequence Numbering

For antibody sequence numbering, it is recommended to run in WeSeq: WeSeq -> Number. Sequence Numbering Introduction Document

Multiple Sequence Alignment

For multiple sequence alignment, it is recommended to run in WeSeq: WeSeq -> Align. Multiple Sequence Alignment Introduction Document

Sequencing Data Analysis

For sequencing data analysis, you can use NGS Analysis. NGS Analysis Introduction Document

Frequency Analysis

For frequency analysis, it is recommended to run in WeSeq: WeSeq -> Frequency. Frequency Analysis Introduction Document

Sequence Mutation

For sequence mutation, it is recommended to operate in WeSeq: WeSeq -> Edit -> Batch Mutate. Alternatively, you can use the Sequence Mutation module.

专利分析包括专利抗体CDR序列搜索、专利序列提取、专利图片OCR。专利分析介绍文档

• 进行专利抗体CDR序列搜索，可以应用模块Patent BLAST。

• 从专利文本文件或专利序列图片OCR提取专利序列，可以应用模块Patent Sequence Listing。

Patent analysis includes searching for antibody CDR sequences in patents, extracting patent sequences, and performing OCR on patent images. Patent Analysis Introduction Document

Patent Antibody CDR Sequence Search

To search for antibody CDR sequences in patents, you can use Patent CDR BLAST.

Extracting Patent Sequences

To extract sequences from patent text files or perform OCR on patent sequence images, you can use Patent Sequence Listing.

• 全原子生成模型BoltzGen
  Binder Design (BoltzGen) 在设计蛋白的同时可以直接生成结构，可设计能够结合各种生物分子靶标的蛋白、肽类等生物分子。另外BoltzGen最厉害和经过验证最多的是用来生成抗体。基于BoltzGen的案例教程

• 从头结构生成
  进行蛋白结构从头生成，可以应用Protein Design (RFDiffusion)。RFDiffusion介绍文档

• 基于主链结构设计序列（逆折叠）

  • ProteinMPNN，建议从WeSeq中运行：WeSeq->Design->ProteinMPNN。ProteinMPNN介绍文档

    

  • ABACUS-R模型，可以使用模块Protein Design (ABACUS-R)。

  • RFDesign模型，可以使用模块Protein Design (RFDesign)。RFDesign介绍文档。

  • ESMIF逆折叠模型，可使用模块Structure Evolution。

• All-atom Generative Model BoltzGen
  Binder Design (BoltzGen) can directly generate structures while designing proteins, capable of designing biomolecules such as proteins and peptides that can bind to various biomolecular targets. Additionally, BoltzGen is most powerful and extensively validated for generating antibodies. Case Tutorial Based on BoltzGen

De Novo Protein Structure Generation

To perform de novo protein structure generation, you can use Protein Design (RFDiffusion). RFDiffusion Introduction Document

Sequence Design Based on Backbone Structure (Inverse Folding)

1. ProteinMPNN

   • To design sequences based on the backbone structure, you can use Protein Design (ProteinMPNN). It is recommended to run it in WeSeq: WeSeq -> Design -> ProteinMPNN. ProteinMPNN Introduction Document

   

2. ABACUS-R

   • You can use Protein Design (ABACUS-R) for sequence design based on backbone structures.

3. RFDesign

   • To use RFDesign for sequence design, you can use Protein Design (RFDesign). RFDesign Introduction Document.

4. ESMIF Inverse Folding Model

   • For sequence design using ESMIF inverse folding model, you can use Structure Evolution.

• BoltzGen
  Binder Design (BoltzGen) 基于全原子生成模型，在设计蛋白的同时可以直接生成结构，可设计能够结合各种生物分子靶标的蛋白、肽类等生物分子。另外BoltzGen最厉害和经过验证最多的是用来生成抗体。基于BoltzGen的案例教程

• RFAntibody
  是基于RFAntibody（抗体微调版RFdiffusion）的抗体从头设计。Antibody Design (RFAntibody)。

• MEAN模型
  基于MEAN模型实现的抗体设计，该模型采用多通道等变图注意力网络，可用于设计CDR的一维序列和三维结构。Antibody Design (MEAN)。

• DiffAb模型
  基于扩散概率模型和等价神经网络的抗体设计，可针对特定抗原结构生成抗体，也可基于抗体-抗原复合物结构进行抗体结构和序列的优化。Antibody Design (DiffAb)。

• BoltzGen
  Binder Design (BoltzGen) is based on an all-atom generative model, can directly generate structures while designing proteins, and is capable of designing biomolecules such as proteins and peptides that can bind to various biomolecular targets. Additionally, BoltzGen is most powerful and extensively validated for generating antibodies. Case Tutorial Based on BoltzGen

• RFAntibody
  RFAntibody is an antibody de novo design method based on the fine-tuned version of RFdiffusion. Antibody Design (RFAntibody).

• MEAN Model
  The MEAN model enables antibody design using a multi-channel equivariant graph attention network, which can be used to design both the one-dimensional sequence and three-dimensional structure of CDRs. Antibody Design (MEAN).

• DiffAb Model
  The DiffAb model utilizes diffusion probabilistic models and equivariant neural networks for antibody design. It can generate antibodies specific to a given antigen structure and optimize antibody structure and sequence based on antibody-antigen complex structures. Antibody Design (DiffAb).

• 利用小样本数据对ESM2蛋白质语言模型进行训练和微调，可使用模块AutoModel Protein

• Train and fine-tune the ESM2 protein language model using small-sample data via AutoModel Protein

• 酶的动力学参数Kcat和Km预测，可使用模块Enzyme Kinetic Prediction。
• 预测酶突变对小分子与蛋白结合能的影响，可使用MD流程，然后使用模块MMGBSA，MMPBSA。
• 酶的同源序列搜索，可使用模块Protein BLAS。
• 酶的逆折叠序列生成，可使用模块Protein Design (LigandMPNN)。
• 酶稳定性改造，可使用流程Protein Stability Optimization。

• For enzyme kinetic parameters Kcat and Km prediction, you can use Enzyme Kinetic Prediction.
• For predicting the effect of enzyme mutations on small molecule-protein binding energy, you can use MD workflow and use MMGBSA，MMPBSA.
• For enzyme homologous sequence search, you can use Protein BLAST.
• For enzyme inverse folding sequence generation, you can use Protein Design (LigandMPNN).
• For enzyme stability engineering, you can use Protein Stability Optimization.

• 预测八种常用蛋白酶的蛋白型裂解位点，包括胰蛋白酶（trypsin）、精氨酸C端肽段（ArgC）、粒胰蛋白酶（chymotrypsin）、谷氨酸C端蛋白酶（GluC）、赖氨酸C端肽段（LysC）、天冬氨酸N端肽段（AspN）、赖氨酸N端肽段（LysN）和L-精氨酸胺基肽酶（LysargiNase），可使用模块Cleavage Site Prediction (DeepDigest)。
• 预测蛋白质序列中潜在的蛋白酶或化学试剂切割位点，可使用模块Cleavage Site Prediction (PeptideCutter)。
• 预测肽段（长度不超过10个氨基酸）被18种基质金属蛋白酶（MMPs）切割的效率及基于指定目标切割谱生成相应的多肽底物，可使用模块Protease (MMP) Cleavage Prediction。

• Predict protein cleavage sites for eight commonly used proteases, including trypsin, ArgC, chymotrypsin, GluC, LysC, AspN, LysN, and LysargiNase, using Cleavage Site Prediction (DeepDigest).
• Predict potential protease or chemical reagent cleavage sites within protein sequences using Cleavage Site Prediction (PeptideCutter).
• Predict the cleavage efficiency of peptides (up to 10 amino acids in length) by 18 matrix metalloproteinases (MMPs), and generate corresponding peptide substrates based on a specified target cleavage profile using Protease (MMP) Cleavage Prediction.

多肽分析包括线性肽/环肽结构预测、多肽对接筛选、线性肽/环肽设计、信号肽预测。

• 进行线性肽结构预测，可以应用模块Peptide Structure Generation。
• 进行环肽结构预测，可以应用模块Cyclic Peptide Structure Prediction。
• 进行基于受体蛋白的多肽对接筛选，可以应用模块Peptide VS。
• 进行环肽设计，推荐应用模块Target-based Peptide Design (EvoBind2)，也可以应用模块Cyclic Peptide Design模块以及模块Protein Design (RFDiffusion) 中的RFPeptide方法。
• 进行线性肽设计，推荐应用模块Target-based Peptide Design (EvoBind2)，也可以应用模块Receptor-Based Peptide Design。

Peptide analysis includes linear/cyclic peptide structure prediction, peptide docking screening, linear/cyclic peptide design, and signal peptide prediction.

• For linear peptide structure prediction, you can use Peptide Structure Generation.
• For cyclic peptide structure prediction, you can use Cyclic Peptide Structure Prediction.
• For receptor protein-based peptide docking screening, you can use Peptide VS.
• For cyclic peptide design, we recommend using Target-based Peptide Design (EvoBind2). You can also use Cyclic Peptide Design and the RFPeptide method in Protein Design (RFDiffusion).
• For linear peptide design, we recommend using Target-based Peptide Design (EvoBind2). You can also use Receptor-Based Peptide Design.

包括密码子优化、CDS优化、UTR优化等。

• 密码子优化，可使用模块Codon Optimization。
• CDS及UTR优化，可使用模块mRNA Optimization (AlphaRNA)。
• UTR优化，可使用模块mRNA 5’UTRs optimization。

Including codon optimization, CDS optimization, UTR optimization, etc.

• For codon optimization, you can use Codon Optimization.
• For CDS and UTR optimization, you can use mRNA Optimization (AlphaRNA).
• For UTR optimization, you can use mRNA 5’UTRs optimization.

靶点鉴定包括疾病相关靶点提取以及小分子靶点预测模块。靶点鉴定介绍文档

• 疾病相关靶点提取，可使用模块Target Prioritization (OpenTargets)，可以提取疾病相关的靶点，支持多种相关性打分。
• 小分子靶点预测，可使用模块Target Prediction (FastTargetPred)，基于二维相似度的小分子靶点预测模块，活性分子及靶点数据来源于ChEMBL数据库。
• 指定基因在肿瘤和正常组织表达情况的检索，可使用模块Tumor Gene Expression (TCGA)，可统计并绘制肿瘤细胞、肿瘤组织、正常组织等的基因表达差异，帮助药物靶点选择、研发立项和决策。

Target identification includes disease-related target extraction and small molecule target prediction modules. Target Identification Introduction Document

• For disease-related target extraction, you can use Target Prioritization (OpenTargets).
• For small molecule target prediction, you can use Target Prediction (FastTargetPred), a small molecule target prediction module based on 2D similarity, with active molecule and target data sourced from the ChEMBL database.
• To search for the expression of specified genes in tumor and normal tissues, you can use Tumor Gene Expression (TCGA). This tool can statistically analyze and visualize the differences in gene expression among tumor cells, tumor tissues, and normal tissues, aiding in drug target selection, research and development project initiation, and decision-making.

小分子生成是从头设计全新分子的过程，可以基于多种AI架构生成类药分子，也可以基于靶点，骨架、活性分子生成衍生物或者相似分子。分子生成介绍文档

• Lead Optimization
  先导化合物优化流程，包括基于阳性分子的结构生成、分子三维相似度筛选、分子对接以及ADMET预测，筛选获得优化分子

• Small Molecule Generation (GenMol)
  基于diffusion model的开源AI框架，用于分子生成。它从大型化学数据库中学习，生成类药物分子。GenMol能够同时优化多种属性（类药物特性、合成可得性），并提供合成规划，大致确保分子可在实验室中合成。

• Small Molecule Generation (EvoMol)
  智能分子生成功能，EvoMol支持全局结构优化与指定子结构的局部优化两种模式，能够在更短时间内批量产生多样性更高、结构新颖性更强的候选分子。

• Small Molecule Generation (REINVENT4)
  基于REINVENT4的小分子生成。支持多种分子生成方式：Reinvent - 从头开始创造新分子，Libinvent - 修饰一个骨架，Linkinvent - 设计两个片段之间的linker，Mol2Mol - 在用户定义的相似度范围内优化分子。

• Scaffold Constrained Small Molecule Generation
  为骨架限制的生成模型，可以限制骨架，指定优化部位，特异性的生成全新分子库。

• Small Molecule Generation from Pocket
  基于DiffSBDD模型实现，可应用受体的结合口袋生成小分子配体。

• Small Molecule Random Generation
  随机类药分子生成模型，基于多种主流的分子生成模型，包括字符级循环神经网络，变分自编码器，以及对抗自编码器的分子生成模块。

Small molecule generation is the process of de novo design of entirely new molecules. It can generate drug-like molecules based on various AI architectures, and can also generate derivatives or similar molecules based on targets, scaffolds, or active molecules. Molecular Generation Documentation

• Lead Optimization
  Lead compound optimization workflow, including structure generation based on positive molecules, molecular 3D similarity screening, molecular docking, and ADMET prediction to screen and obtain optimized molecules.

• Small Molecule Generation (GenMol)
  An open-source AI framework based on diffusion models for molecular generation. It learns from large chemical databases to generate drug-like molecules. GenMol can simultaneously optimize multiple properties (drug-likeness, synthetic accessibility) and provides synthetic planning to roughly ensure molecules can be synthesized in the laboratory.

• Small Molecule Generation (EvoMol)
  Intelligent molecular generation function. EvoMol supports two modes: global structure optimization and local optimization of specified substructures, capable of generating candidate molecules with higher diversity and stronger structural novelty in shorter time.

• Small Molecule Generation (REINVENT4)
  Small molecule generation based on REINVENT4. Supports multiple molecular generation methods: Reinvent - create new molecules from scratch, Libinvent - modify a scaffold, Linkinvent - design linkers between two fragments, Mol2Mol - optimize molecules within user-defined similarity ranges.

• Scaffold Constrained Small Molecule Generation
  A scaffold-constrained generation model that can restrict scaffolds, specify optimization sites, and specifically generate entirely new molecular libraries.

• Small Molecule Generation from Pocket
  Based on the DiffSBDD model, it can generate small molecule ligands using the receptor’s binding pocket.

• Small Molecule Random Generation
  Random drug-like molecule generation model based on various mainstream molecular generation models, including character-level recurrent neural networks, variational autoencoders, and adversarial autoencoder molecular generation modules.

虚拟筛选根据配体或受体结构，对小分子化合物进行筛选，预测可能的活性分子，大大提高化合物药物发现进程，缩减药物发现费用。

基于受体的方法

• 虚拟筛选流程Cascaded Virtual Screening，基于配体和基于受体整合的虚拟筛选流程。

基于配体方法

• 性质过滤
  • Property Filter，根据性质过滤，包括基于相似度（2D和3D）,基于性质过滤，基于结构聚类，基于对接等。Property Filter介绍文档
  • PAINS Filter，过滤PAINS片段。PAINS Filter介绍文档
• 结构搜索
  • Substructure Search，从化合物库中查找含有特定子结构片段的化合物。
  • 2D Similarity Search，从化合物库中查找出与查询分子二维相似的化合物。二维形状搜索介绍文档
• 三维形状搜索
  • AlphaShape，从化合物库中查找出与查询分子三维形状相似的化合物，私有库需要结合 3D Conf Generation (AlphaConf) 模块生成筛序库的分子构象。三维形状搜索介绍文档
• 结构聚类
  • Diverse Subset，通过分子聚类，挑选具有代表性的结构多样性的分子子集，常用于基于其他筛选手段得到的分子库进行进一步挑选，减小筛选hits的数量。

Virtual screening is a computational technique used to identify potential active compounds by screening large libraries of small molecules. This process can significantly accelerate drug discovery and reduce costs.

Receptor-Based Methods

1. Virtual Screening Workflow
   • Cascaded Virtual Screening: An integrated virtual screening workflow that combines ligand-based and receptor-based methods.

Ligand-Based Methods

1. Property Filtering

   • Property Filter: Filters compounds based on properties such as 2D and 3D similarity, property-based filtering, structural clustering, and docking. Property Filter Introduction Document
   • PAINS Filter: Filters out compounds containing PAINS fragments. PAINS Filter Introduction Document

2. Structure Search

   • Substructure Search: Searches for compounds containing specific substructures within a compound library.
   • 2D Similarity Search: Finds compounds in a library that are 2D similar to the query molecule. 2D Similarity Search Introduction Document

3. 3D Shape Search

   • AlphaShape: Searches for compounds in a library that have a 3D shape similar to the query molecule. For private libraries, it needs to be combined with the 3D Conf Generation (AlphaConf) module to generate molecular conformations for screening. 3D Shape Search Introduction Document

4. Structure Clustering

   • Diverse Subset: Selects a representative subset of structurally diverse molecules through clustering. This is often used to further refine a set of hits obtained from other screening methods, reducing the number of hits.

分子性质包括小分子的理化性质以及药代动力学（ADMET）性质。

• 理化性质计算，建议使用Descriptors (RDKit)。理化性质计算介绍文档
• 化合物可合成性评估，建议使用Synthetic Accessibility Score。化合物可合成性评估介绍文档
• P450代谢位点预测，建议使用Metabolism Site Prediction。P450代谢位点预测介绍文档
• 毒效片段识别，建议使用Toxic Fragment Identification。毒效片段识别介绍文档
• ADMET性质预测，建议使用ADMET Prediction。ADMET介绍文档

Molecular properties include the physicochemical properties and pharmacokinetic (ADMET) properties of small molecules.

• Physicochemical Properties Calculation, recommended tool: Descriptors (RDKit). Physicochemical Properties Calculation Documentation
• Synthetic Accessibility Evaluation, recommended tool: Synthetic Accessibility Score. Synthetic Accessibility Evaluation Documentation
• P450 Metabolism Site Prediction, recommended tool: Metabolism Site Prediction. P450 Metabolism Site Prediction Documentation
• Toxic Fragment Identification, recommended tool: Toxic Fragment Identification. Toxic Fragment Identification Documentation
• ADMET Properties Prediction, recommended tool: ADMET Prediction. ADMET Prediction Documentation

分子对接是研究相互作用的重要工具，包括蛋白-小分子，蛋白-蛋白对接。

• 蛋白-小分子对接

  • Molecular Docking (AutoDock-GPU)，建议从WeView中运行：WeView->Docking。基于GPU加速的AutoDock的分子对接工具。AutoDock-GPU对接介绍文档

    

  • Molecular Docking (SMINA)，基于Autodock Vina分支SMINA的分子对接工具。SMINA对接介绍文档

  • Molecular Docking (DiffDock)，基于扩散生成模型的对接工具。DiffDock对接介绍文档

  • Molecular Docking (Gnina)，基于深度学习的分子对接工具，采用卷积神经网络（CNN）评分函数对配体-受体结合构象进行打分和排序。

• 蛋白-蛋白/核酸对接

  • Protein Docking (HDOCK)，支持蛋白-蛋白，蛋白-DNA/RNA对接，支持限制位点对接。
  • Protein Docking (FRODOCK)，除了常规蛋白-蛋白，还支持抗体抗原模式对接，支持限制位点对接。
  • Antibody-Antigen Docking (HADDOCK)，支持抗体抗原模式对接。

Molecular docking is an important tool for studying interactions, including protein-small molecule and protein-protein docking.

• Protein-Small Molecule Docking

  • Molecular Docking (AutoDock-GPU), recommended to run from WeView: WeView->Docking. A GPU-accelerated molecular docking tool based on AutoDock. AutoDock-GPU Docking Documentation

    

  • Molecular Docking (SMINA), a molecular docking tool based on the AutoDock Vina branch SMINA. SMINA Docking Documentation

  • Molecular Docking (DiffDock), a docking tool based on diffusion generative models. DiffDock Docking Documentation

  • Molecular Docking (Gnina), A deep learning-based molecular docking tool that employs convolutional neural network (CNN) scoring functions to score and rank ligand-receptor binding poses.

• Protein-Protein/Nucleic Acid Docking

  • Protein Docking (HDOCK), supports protein-protein, protein-DNA/RNA docking, and site-specific docking.
  • Protein Docking (FRODOCK), supports not only conventional protein-protein docking but also antibody-antigen mode docking and site-specific docking.
  • Antibody-Antigen Docking (HADDOCK), supports antibody-antigen mode docking.

分子格式转换工具，包括不同格式文件转换、氨基酸字母格式转换等。

• 基于Open Babel的分子文件格式转换工具，可以使用Format Conversion (Open Babel)。
• 基于RDKit的分子文件格式转换工具，可以使用Format Conversion (RDKit)。
• 小分子构象生成工具3D Conf Generation (AlphaConf)，可将生成的二进制构象压缩文件AC.GZ转为便于查看的SDF文件。
• 氨基酸格式转换工具，可使用3-letter AA Conversion，将氨基酸缩写三字母格式转换为单字母格式。

Molecular format conversion tools, including conversion of different format files, amino acid letter format conversion, etc.

• For molecular file format conversion based on Open Babel, you can use Format Conversion (Open Babel).
• For molecular file format conversion based on RDKit, you can use Format Conversion (RDKit).
• For small molecule conformation generation, use the tool 3D Conf Generation (AlphaConf), which can convert the generated binary conformation compressed file AC.GZ to an SDF file for easy viewing.
• For amino acid format conversion, you can use 3-letter AA Conversionto convert amino acid abbreviations from three-letter format to one-letter format.

轨迹分析对分子动力学模拟后产生的轨迹进行结构分析，观察研究对象在模拟过程中的动态变化。

教程

分子动力学介绍文档

Trajectory analysis involves structural analysis of the trajectories generated from molecular dynamics simulations to observe the dynamic changes of the study object during the simulation process.

Tutorial

Molecular Dynamics Introduction Documentation

结合自由能计算是预测分子间结合强弱的重要方法。

教程

分子动力学介绍文档

Combining free energy calculations is a crucial method for predicting the strength of molecular interactions.

Tutorial

Molecular Dynamics Introduction Documentation