GeneVis: 3D Gene Expression Visualization

• 

  Only Png, Demo in the Github Links!

Inspiration

Modern molecular biology and genetics research heavily relies on large-scale datasets. Proper preprocessing and visualization of such data are crucial for accelerating biomedical discoveries. The inspiration behind this project was to create a tool that makes gene expression data more accessible, interpretable, and visually insightful.

What I Learned

Through this project, I learned to:

Apply preprocessing techniques on high-dimensional genomic datasets,

Use Principal Component Analysis (PCA) for dimensionality reduction,

Implement 3D visualization methods,

Leverage Python’s data science ecosystem (Pandas, NumPy, Scikit-learn),

Build interactive bioinformatics applications using Streamlit.

How I Built It

The project was developed in the following steps:

Data Preprocessing: Cleaning missing values and transforming the raw dataset into an analysis-ready format.

Dimensionality Reduction: Reducing the gene expression dataset (with over 54,000 features) into 3 dimensions using PCA.

𝑍=𝑋𝑊,𝑊∈𝑅𝑛×3Z=XW,W∈Rn×3

Visualization: Generating 3D PCA scatter plots using Plotly.

Interactive Interface: Deploying a user-friendly application with Streamlit.

Challenges

The dataset was extremely high-dimensional (50,000+ features), which posed challenges for memory management and computation time.

Optimizing PCA required the use of chunking strategies (processing the dataset in smaller parts).

Designing a 3D visualization that was both academically rigorous and user-friendly required iterative refinement.

Conclusion

This project demonstrates an effective pipeline for preprocessing, analyzing, and visualizing high-dimensional gene expression data. The methodology can be easily extended to other types of biological datasets and disease-focused research, offering a foundation for future bioinformatics applications.

Built With

• kaleido
• numpy
• pandas
• plotly
• python
• scikit-learn
• streamlit