Day One| Digitalization and Basic Characterization of Cell Data

Morning: Theoretical Explanation (Phase I & II)

Phase One: Cellular Data Digitization (Data Representation）

Core Objective:Solve"How to Make Cells Understandable to AI"
• The complexity of cellular multi-omics data (RNA, ATAC, Protein, Spatial)
• Best Practices for Data Standardization and Quality Control
• The core logic from raw data to machine-readable structure

Supporting Model Theory:
• MultiVI: Unified Representation of RNA+ATAC Multimodalities (Key Explanation)
• totalVI: RNA+Protein Joint Encoding
• MOFA+: Multi-Omics Factor Analysis
• OmniReg-GPT (New Model)，NC2026）：Basic Characterization of DNA Sequences, Identification of Genomic Loci and Expression Prediction

Stage Two: Cell State Modeling (State Learning）

Core Objective:Solve"How to Identify the State of a Cell"
• The Logic of Transformation from "Cell Data" to "Cell State"
• Biological Significance of Latent Variable Space
• Identification of Cell Subsets and Discovery of Rare Cells

Supporting Model Theory:
• scVI/scANVI: Single-cell latent variable modeling (core)
• β-VAE: Disentangled Representation Learning
• Contrastive Cell Embedding: Application of Contrastive Learning in Cell Representation

Afternoon: Practical Exercises (Corresponding to Theoretical Stages 1 and 2 from the Morning)

Preparation for Practical Operations:GPU Server Environment Adaptation, Linux and Python Environment Debugging

1. Advanced Common Linux Commands:Cell Data File (Single CellBatch management, permission settings, and format conversion of RNA and ATAC data;
2. Python Environment Setup and Optimization:Cell Data Processing Related Package (Installation and debugging of scanpy, torch, scvi-tools.

Practical Model Explanation (Python Code Parsing + Hands-on Practice on GPU Server

1. Practical Model 1: MultiVI (Multimodal Unified Representation) ——Corresponding to the first phase theory, achieveUnified encoding of RNA+ATAC data, completed data denoising and batch effect correction, mastered the construction method of latent variable space, and understood its core role as the model base;

2. Practical Model 2: scVI (Single-cell Latent Variable Modeling) ——Corresponding to the first and second phase theories, based on single-cellRNA data, completed latent variable modeling, preliminary analysis of cell clustering, mastered the training and evaluation methods of basic representation models, and met the core requirements for cell state identification;

3. Practical Model:OmniReg-GPT Demo (New Model)—— DNA Sequence Feature Extraction, Gene Expression Prediction, Understanding the Application of Basic Representation Models in Genomics, and Showcasing the Core Technology of a Nature Communications Paper.

The Next Day| Cell State Modeling and Spatial Transcriptomics

Morning: Theoretical Explanation (Deepening of the Second Phase)

Fundamentals of Spatial Transcriptomics

Core Objective:Solve"Spatial State of Cells in Tissue"
• Overview of Spatial Transcriptomics Technologies (Visium, Stereo-seq, MERFISH)
• Cell State Identification under Spatial Constraints
• Tissue Microenvironment and Cell Communication

Supporting Model Theory:
• GraphST: Graph Neural Network Spatial Representation
• STALigner: Spatial Transcriptomics Cross-Sample Integration

• Nicheformer(New model,2025NM）: SpaceBase Model

Afternoon: Practical Exercises (Corresponding to Morning Spatial Transcriptomics Theory)

Preparation for Practical Operations: Preprocessing of Spatial Transcriptomics Data and Toolkit Debugging

1. Python Package Adaptation:Installation and debugging of PyTorch Geometric (Graph Neural Network) and squidpy (spatial analysis) toolkits;
2. Data Preprocessing Review:Spatial transcriptomics data format (Visium, Stereo-seq) read and preprocessing methods.

Practical Model Explanation (Python Code Parsing + Hands-on Practice on GPU Servers

1.Practical Model:GraphST Practical Application (Spatial Data Clustering and Domain Recognition) ——Based on spatial transcriptomics data, construct spatial graph networks, complete tissue domain identification and spatial clustering, and master the application of graph neural networks in spatial data;

2. Practical Model:STAlignerPractical Operation (Spatial Transcriptomics Cross-sample Integration—— Understand how to eliminate batch effects in spatial transcriptomics, master the basic principles and core methods of batch removal, and comprehend the modeling approach for spatial data.
3. Practical Model: Nicheformer Practice (Spatial Base Model) ——Characterization of the cellular microenvironment, mastering the core applications of spatial foundational models, and enhancing practical skills in cellular state identification.

Day Three| Regulatory Mechanism Inference and Cell Dynamics Prediction

Morning: Theoretical Explanation (Third and Fourth Stages)

Stage Three: Modeling of Cellular Regulation Mechanisms (Regulatory Modeling）

Core Objective:Solve"Why Do Cells Change?"
• The underlying mechanism of cell regulation
• From phenotypic recognition to mechanism level
• The Core Value of Regulatory Mechanism Modeling in Drug Development

Supporting Model Theory:
• GAT: Graph Attention Network, Gene Regulatory Network Inference
• SCENIC: Inference of Transcription Factor Regulation
• Gene Regulatory Graph: Causal Relationship Modeling

Stage Four: Cellular Dynamics Prediction (Dynamic Evolution）

Core Objective:Solve"Where Will Cells Go Next?"
• The Core Logic of Cell Fate Trajectory Inference
• The Importance of Dynamic Prediction in Drug Development (e.g., Drug Resistance, Relapse Prediction)

Supporting Model Theory:
• CellRank 2: Fate Probability and Trajectory Inference
• RNA Velocity: Transcription Dynamics Modeling
• stVCR (New Model, Nat Methods 2026): Spatial cell developmental trajectory inference, based onModeling of Spatial-Gene Dual Velocity Field in Neural ODE

Afternoon: Practical Drills (Corresponding to the Third and Fourth Stages of Theory in the Morning)

Preparation for Practical Operations: Debugging of Graph Neural Networks and Dynamic Prediction Toolkits

1. Python Package Adaptation:Installation and debugging of PyTorch Geometric (Graph Neural Network) and CellRank (dynamic prediction) toolkits;
2. Data Preprocessing Review:Review the preprocessing methods for gene expression data and regulatory relationship data related to theoretical content from this morning.

Practical Model Explanation (Python Code Parsing + GPU Server Hands-on Practice

1.Practical Model:SCENIC(Regulatory Network Mechanism Inference)—— Corresponding to the third phase theory, construct gene regulatory networks based on gene expression data, identify key regulatory nodes, master the core methods of mechanistic reasoning, and understand their applications in drug target discovery.

2. Practical Model: CellRank 2 (Fate and Trajectory Inference) ——Corresponding to the fourth stage theory, based on single-cell data, deduce cell differentiation trajectories, predict future cell states, master core methods of dynamic prediction, and meet the needs of drug resistance and recurrence prediction in drug development.

3. Practical Model: stVCR Practical (New Model) ——Spatial trajectory inference, predict cell differentiation direction, understandCore Principle of Neural ODE Modeling Spatial-Gene Dual Velocity Field, Showcasing Key Technology from Nature Methods 2026 Paper;

Day Four| Drug Perturbation Modeling and Disease Systems

Morning: Theory Explanation (Fifth and Sixth Stages)

Phase Five: Drug Action Modeling (Drug Perturbation Modeling）

Core Objective:Solve"How Drugs Alter Cell Fate"
• The Core Logic of Drug Action on Cells
• Application Scenarios of Drug Perturbation Modeling in the Entire Drug Development Process

Supporting Model Theory:
• ChemCPA: Drug Dose-Response Modeling
• scGen: Perturbation Response Generation
• CellOT: Optimal Transport Perturbation Prediction
• scGPT：Large Model Prediction Perturbation

Stage Six: Disease System Modeling (Disease System Modeling）

Core Objective:Solve"How Cell Networks Are Reconstructed in Disease"
• Patterns of Cellular Network Changes in Disease States
• The Core Value of Disease System Modeling in Patient Stratification and Disease Subtype Prediction

Supporting Model Theory:
• DeepProg: Disease Prognosis Prediction
• Numbat-multiome: Inferring CNV from Single-cell Multi-omics Data and Reconstructing Tumor Phylogeny

Afternoon: Practical Exercises (Corresponding to the Fifth and Sixth Stages of Theory in the Morning)

Preparation for Practical Operation: Debugging of Drug Perturbation Model Toolkit

1. Python Package Adaptation:Installation and debugging of drug perturbation-related toolkits such as ChemCPA and scGen;
2. Data Preparation:Preprocessing and Import Methods for Drug-Related Data (Drug Dosage, Cellular Response Data).

Practical Model Explanation (Python Code Parsing + Hands-on Practice on GPU Server

1. Practical Model: ChemCPA (Drug Perturbation Prediction) ——Corresponding to the fifth stage theory, construct a drug perturbation model to predict the effects of different drug doses, combination drug responses, master the core capabilities of virtual screening, and understand its role in drug development.The Role in ROI Improvement;

2. Practical Model: scGen Practical Operation (Single-Drug Perturbation Response Generation) ——Based on single-cell data, generate predictions of cellular states after drug perturbations and master the core methods of generative perturbation models;

3. Practical OperationModel：DeepProg（DiseasePrognosisAnalysis）——Based on Multi-Omics Data andAIModel, analyze disease statesPatient Prognosis Progress。

Day Five| Digital Twin and Virtual Clinical Applications

Morning: Theoretical Explanation (Seventh and Eighth Stages)

Stage Seven: Digital Twin Cells/Tissue (Digital Twin)

Core Objective:Solve"How to Build a Deductive Virtual Human Local System"
• Application Logic of Digital Twin Technology at the Cellular and Tissue Levels
• Its core value in reducing pharmaceutical companies' wet lab costs

Supporting Model Theory:

• Virtual cell: VirtualCell Overview
• DrugCell: Drug Response Neural Network
•PhysiCell（Cell 2026）：Cell Simulation Engine

Stage Eight: Virtual Clinical Trials and Drug Development (Virtual Clinical Translation）

Core Objective:Solve"How to Directly Serve Drug Development and Clinical Decision-Making"
• The Design Logic of Virtual Clinical Trials
• Prediction Chain from In Vitro to In Vivo
• ROI Calculation and Decision Support

Supporting Model Theory:
• PK/PD Neural Surrogate: Pharmacokinetic Neural Network
• Clinical Response Simulator: Clinical Response Simulator

Afternoon: Practical Exercises+ Course Summary

Preparation for Practical Operations: Debugging of Digital Twin and Virtual Clinical Model Toolkits

1. Python Package Adaptation:Installation and debugging of digital twin-related toolkits such as DrugCell and PhysiCell.

Practical Model Explanation (Python Code Parsing + GPU Server Hands-on Practice

1. Practical Model: DrugCell (Industry-Level Drug Response Prediction) ——Corresponding to the seventh stage theory, construct a drug response prediction model, explain the mechanism of drug action, master the application methods of industry-level models, and understand their role in reducing wet-lab experimental costs.

2. Practical Model: PhysiCell (Digital Twin Underlying Simulation) ——Corresponding to the seventh-stage theory, a virtual cell simulation environment is built to complete the simulation closed-loop from virtual cells to virtual tissues, mastering the underlying operations of digital twins and connecting to virtual clinical applications;

Course Summary

• Technology Stack Review:From the data→ Status → Regulation → Dynamics → Drugs → Diseases → Twinning → Clinical
• Frontier Trends：Large Models, Multimodal, Spatial Omics, Virtual Knockout
• Career Development：Computational Biology Talent Demand and Competency Pathways

Supporting Resources

• Course PPT (Theoretical Explanation)
• Practical Code Package (Jupyter Notebook)
• GPU Server Account (Cloud Practical Operation)
• Dataset (Public Single-cell/Spatial Data)
• References (Latest Top Journal Articles)，Basically, it isNew Articles for 2026 and 2025 + A Few Classic Articles）

Day One

1. Overview of Genome Editing Technology

1.1 A Brief Overview of the Era of Genomic Sequencing, Editing, Reading and Writing, and the Current Status of Genome Editing Technology

2. Principles of Four Generations of Genome Editing Technology

2.1 The Development Process of the Fourth-Generation Genome Editing Technology

2.2 Composition of ZFN, TALEN, and CRISPR/Cas SystemsWorkPrinciple

3. Origin and Classification of CRISPR/Cas Systems

31. CRISPR/Cas SystemTheDiscovery Process

3.2 CRISPR/Cas SystemThe Principle of Adaptive Immunity

3.3 Classification Basis and Types of CRISPR/Cas Systems

4. CRISPR/Cas System-Mediated DNA Editing Tool

4.1 CRISPR/Cas9 Gene Editing Tool

4.2 CRISPR/Cas12a Gene Editing Tool

5. Development of CRISPR/Cas System-Derived Tools

5.1 Composition, Mechanism of Action, and Applications of Base Editing Tools

5.2 Prime EditingMechanism of Action, Applications, and Development Trends

6. CRISPR/Cas-Mediated Gene Regulation, Cell Imaging, and Nucleic Acid Detection Technologies

6.1 CRISPR/CasMediatePrinciples and Tool Components of Gene Regulation Technology

6.2 CRISPR/CasMediatePrinciples and Tool Composition of Cell Imaging Technology

6.3 CRISPR/CasMediatePrinciples and Tool Composition of Nucleic Acid Testing Technology

The Next Day

1. Off-target Effects and Their Detection

1.1 Methods for detecting off-target effects: Amplicon sequencing, Whole genome sequencing,GUIDE-seq, etc.

1.2 Methods to Avoid Off-Target Effects

2. Gene Editing Process- Taking plants as an example

2.1 Target SitesgRNAOrcrRNADesign Principles

2.2 Methods for Designing and Constructing Expression Cassettes

2.3 Plant Protoplast Transient Expression System

2.4 Genetic Transformation of Gene Editing Vectors

2.5 Detection of Gene-Edited Mutants

3. Hands-on Practice with Commonly Used Software for Genome Editing

3.1 Target Site Design SoftwareCas-Designer、BE-Designer、PE-Designeretc.

3.2 Mutation Analysis SoftwareCas- Analyzer、BE-Analyzer、PE- Analyzer

4. Current Status and Prospects of Genome Editing Technology Applications in Various Fields

4.1 Applications of Genome Editing Technology in Gene Therapy, Immunology, and Virus Diagnosis

Day Three Theoretical Part (Artificial Intelligence+Gene Editing Background)

1. Overview of Deep Learning

1.1. The Foundation of Deep Learning

1.2. How Deep Neuronal Networks Work

1.3. Development Trends of Deep Learning Technology:Advances in Self-Supervised Learning, Transfer Learning, and Few-Shot Learning

2. Application of Deep Learning in Gene Editing

2.1. Applications Based on Supervised Learning: Sequence Labeling Model

2.2. Application of Zero-Shot Prediction Models: Structural Models, Large Language Models, Multimodal Models,

2.3. Application of Few-shot Prediction Framework (Design-Build-Test-Learn and Lab-in-the-loop Paradigm)

3. Deep Learning ingRNAApplication in Optimization and Design

3.1. gRNA Activity Prediction

3.2. Off-target Effect Prediction

3.3. Introduction to gRNA Prediction Model

4. AIApplications of Auxiliary Protein Directed Evolution in Gene Editing

4.1. Basic Concepts and Experimental Methods of Protein Directed Evolution

4.2 AI-Assisted Protein Evolution Tools

4.3. Integration of AI and Experimental Feedback

5. AIApplication of Protein Design in Gene Editing

5.1. Protein Design Tools

5.2. Enzyme Design

5.3. Binder Design

6. Application of AI Enzyme Mining in Gene Editing

6.1. Mining Gene Editing Enzymes Based on Large Language Models

6.2. Mining Gene Editing Enzymes Based on Structural Alignment

Day FourHands-on Teaching of Deep Learning Applications in Gene Editing

1. Basic Knowledge and Environment Setup

1.1. GPU Server Login

1.2. Linux Basics

1.3. Python Basics

1.4. Introduction and Installation of Common Deep Learning Toolkits

2. Using Deep Learning to PredictgRNAActivity

2.1. Set Up Deep Learning Environment，InstallationTools Required for gRNA Activity Prediction

2.2. High-Throughput Data Acquisition:Introduction and Use of Public Datasets

2.3. Dataset Division:Training set, validation set, test set

2.4. Model Construction and Debugging:Deep Learning Model Architecture Design (e.g.CNN, RNN）

2.5. Model Performance Evaluation:Precision, Recall,F1 Score and Other Evaluation Metrics

2.6. gRNA Activity Prediction: Demonstration of Practical Application Cases and Interpretation and Application of Prediction Results

3. Predicting Editing Activity Using Deep Learning

3.1. Environment Configuration:Install the required tools and libraries

3.2. Data Acquisition:Edit Activity-Related DatasetsCleaning

3.3. Dataset Division

3.4. Model Construction and Debugging

3.5. Model Performance Evaluation

3.6. Editing Activity Prediction:Presentation and Interpretation of Prediction Results

4. Zero-shot Protein Evolution ToolAiCEPractical Operation

4.1. Principle and Application Scenarios of AiCE

4.2. Environment Setup

4.3. Use of the Inverse Folding Model: How to Utilize AiCE for High-Activity Mutation Prediction; Case Demonstration and Practical Operation

4.4. Application Examples: Efficient Evolution of Base Editors

5. Few-shot Protein-directed Evolution ToolEVOLVEproPractical Operation

5.1. Background and Application of EVOLVEpro

5.2. Environment Setup and Configuration

5.3. Few-shot Fine-tuning Based on DMS Data

5.4. Few-shot Fine-tuning Based on Experimental Data Feedback

5.5. Application Example: Efficient Evolution of Cas12f

Day Five Case Reproduction of Gene Editing Tool Design and Mining

1. Design MLH1 Binder Enhances Prime Editing (PE) Efficiency

1.1. Background Knowledge: R-basedFdiffusion + ProteinMPNN + AlphaFoldTheBinder Design Process

1.2. Environment Setup and Configuration

1.3. Input Structure Preparation (AlphaFold Prediction)

1.4. Structure Backbone Generation: Utilizing RFdiffusion for structural sampling and optimization to generate protein structure backbones

1.5. Sequence Design: Optimize sequence design based on the structural backbone generated by RFdiffusion.

1.6. Complex Structure Prediction Validation: Using AlphaFold for structural prediction of the binder and target protein complex,Verify whether the designed complex structure meets expectations

1.7. Result Visualization:UseVisualization of Structures and Design Results Using PyMOL

2. Cas13Inhibitor Design

2.1. Background Knowledge:Introduction to the Structure and Function of Cas13

2.2. Input Structure Preparation

2.3. Protein Design Process:CombinationRFdiffusion, ProteinMPNN and AlphaFold Design Cas13 Inhibitors

2.4. Design Result Analysis and Visualization

3. Mining Novel Based on Protein Language ModelsCRISPRSystem

3.1. Introduction and Process of Protein Language Models in Enzyme Mining

3.2. Introduction to Sequence Databases and Download

3.3. Query Sequence Preparation

3.4. Mining Cas12 Family Gene Editing Enzymes Based on ESM Language Model

4. Novel Mining Based on 3D StructureCRISPR System

4.1. Background Knowledge of Structural Alignment: Importance and Applications of Structural Alignment; Comparison of the Advantages and Disadvantages of Different Structural Alignment Tools

4.2. Introduction to the Foldseek Series of Tools:IntroductionFoldseek、Foldseek multimer, Folddisco, FoldMason, etc.Basic Principles and Use of Tools

4.3. Introduction and Download of Structural Databases: PDB, AFDB, ESM Atlas

4.4. Input Structure Preparation:Prepare the target protein structure file for comparison

4.5. Using Foldseek Web Version:Demonstration of how to useFoldseek Web Version for Structural Alignment；Explain How to Interpret the Output Results and Conduct Follow-up Analysis

4.6. Using Foldseek Local Version:Local DeploymentFoldseek and use the command line tool for alignment

4.7. Using Local Versions of DALI and TM-align Tools: Introduction to DALI and TM-align ToolsInstallation and Use of the Local Version

4.8. Construction of Structural Evolutionary Tree:UseFoldMason Builds Evolutionary Trees of Protein Structures

Day One

Multi-omics Sequencing Technology and DatabasesMorningTheoretical Explanation

Day One

1.Overview of AIDD and Introduction to Comprehensive Drug Database

2. Overview of Artificial Intelligence-Aided Drug Design (AIDD)

3. Installation Environment

(1)anaconda

(2)vscode

(3)pycharm

(4) Virtual Environment

4. Basic Usage of Third-Party Libraries

(1)numpy

(2)pandas

(3)matplotlib

(4)requests

5. Acquisition Methods of Comprehensive Databases for Multiple Drugs

(1) KEGG (requests crawler)

(2)Chebi（libChEBIpy）

(3)PubChem（pubchempy / requests）

(4)ChEMBL（chembl_webresource_client）

(5)BiGG（curl）

(6)PDB（pypdb）

The Next Day

2. ML-based AIDD

1. Machine Learning

(1) Types of Machine Learning:

①Supervised Learning

②Unsupervised Learning

③ Reinforcement Learning

(2) Typical Machine Learning Methods

① Decision Tree

② Support Vector Machine

③Naive Bayes

④ Neural Network

⑤Convolutional Neural Network

(3) Evaluation and Validation of the Model

(4) Classification Evaluation: Accuracy, Precision, Recall, F1 Score, ROC Curve, AUC Calculation

(5) Regression Evaluation: Mean Absolute Error, Mean Squared Error, R2 Score, Explained Variance Score

(6) Cross-validation

2. Basic Usage of sklearn Toolkit

3. Basic Usage of RDKit Toolkit

4. Compound Encoding Methods and Compound Similarity Theoretical Knowledge

5. Project Practice 1: Molecular Screening Based on ADME and Ro5

6. Project Practice 2: Ligand Screening Based on Compound Similarity

7. Project Practice 3: Molecular Clustering Based on Compound Similarity

8. Project Practice 4: Machine Learning-Based Bioactivity Prediction

9. Project Practice 5: Machine Learning-Based Molecular Toxicity Prediction

Day Three

3. GNN-based AIDD

1. Graph Neural Network

(1) Framework Introduction: PyG, DGL, TorchDrug

(2) Graph Neural Network Message Passing Mechanism

(3) Design of Graph Neural Network Datasets

(4) Practical Tasks of Graph Neural Network: Node Prediction, Graph Prediction, and Edge Prediction

2. In-depth Analysis of the Paper: DeepTox: Toxicity Prediction using Deep Learning

3. Project Practice 1: Molecular Toxicity Prediction Based on Graph Neural Networks

(1) Construction of PyG Graph Dataset from SMILES Molecular Data

(2) Molecular Toxicity Prediction Based on GNN

4. Project Practice 2: Prediction of Protein-Ligand Interactions Based on Graph Neural Networks

(1) Protein molecule visualization, constructing PyG graph dataset

(2) Network construction and interaction prediction based on GIN

Day Four

4. NLP-based AIDD

1. Natural Language Processing

(1) Encoder-Decoder Model

(2) Recurrent Neural Network RNN

(3)Seq2seq

(4)Attention

(5)Transformer

2.Project Practice 1: Molecular Toxicity Prediction Based on Natural Language

(1)SMILES Molecular Dataset Word Vector Representation Method

(2) Molecular Toxicity Prediction Based on NLP Model

3. Project Practice 2: Prediction of Organic Chemical Reaction Yields Based on Transformer (Prediction of chemical reaction yields using deep learning)

4. In-depth Paper Reading and Code Explanation: "Mapping the space of chemical reactions using attention-based neural networks"

Day Five

5. Molecular Generation and Drug Design

1. Molecular Generation Model

(1) Recurrent Neural Network (RNN)

(2) Variational Autoencoder VAE

(3) Generative Adversarial Network (GAN)

(4) Reinforcement Learning (RL)

2.Project Practice 1: Small Molecule Compound Generation Model Based on Graph Data "A Graph to Graphs Framework for Retrosynthesis Prediction"

3. Project Practice 2: NLP-Based Antibody Generation Model "Generative Language Modeling for Antibody Design"

Introduction of the Lecturer